3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-279.2.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-279.2.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
22 Severity : enhancement
24 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
26 Severity : enhancement
28 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
30 Severity : enhancement
32 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
34 --------------------------------------------------------------------------------
36 04-30-2012 Whamcloud, Inc.
38 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
39 * Server support for kernels:
40 2.6.32-220.4.2.el6 (RHEL6)
41 * Client support for unpatched kernels:
42 2.6.32-220.4.2.el6 (RHEL6)
44 2.6.18-238.19.1.el5 (RHEL5)
45 2.6.16.60-0.69.1 (SLES10)
46 * Recommended e2fsprogs version: 1.41.90.wc4
47 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
48 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
49 stack overflow. For more information, please refer to bugzilla 17630.
50 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
51 the re-exporting nodes to be mounted with "32bitapi" mount option
53 Severity : enhancement
55 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
57 Severity : enhancement
59 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
61 Severity : enhancement
63 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
65 Severity : enhancement
67 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
69 --------------------------------------------------------------------------------
71 09-30-2011 Whamcloud, Inc.
73 * Server support for kernels:
74 2.6.18-238.19.1.el5 (RHEL5)
75 2.6.32-131.6.1.el6 (RHEL6)
76 * Client support for unpatched kernels:
77 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
78 2.6.18-238.19.1.el5 (RHEL5)
79 2.6.32-131.6.1.el6 (RHEL6)
80 2.6.29.4-167.fc11 (FC11)
81 2.6.16.60-0.69.1 (SLES10)
83 * Recommended e2fsprogs version: 1.41.90.wc3
84 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
85 Lustre filesystem with 4K stack may cause a stack overflow. For more
86 information, please refer to bugzilla 17630.
87 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
88 mounted with "32bitapi" mount option
90 Severity : enhancement
91 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
93 Severity : enhancement
94 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
96 Severity : enhancement
97 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
100 Frequency : only with SLES10
102 Description: Use OFED "KMP" provided by Novell
103 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
104 Lustre is now built against this package. That means you need to
105 install the ofed-kmp package from Novell for the patchless client.
106 Note that the ofed-kmp that Novell ships may not exactly match the
107 kernel version but should still be compatible.
109 Severity : enhancement
111 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
112 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
114 --------------------------------------------------------------------------------
116 2010-07-15 Oracle, Inc.
118 * Server support for kernels:
119 2.6.18-164.11.1.el5 (RHEL 5)
120 2.6.18-164.11.1.0.1.el5 (OEL 5)
121 * Client support for unpatched kernels:
122 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
123 2.6.18-164.11.1.el5 (RHEL 5),
124 2.6.18-164.11.1.0.1.el5 (OEL 5)
125 2.6.16.60-0.42.8 (SLES 10),
127 2.6.29.4-167.fc11 (FC11)
128 * Recommended e2fsprogs version: 1.41.10-sun2
130 Severity : enhancement
132 Description: Distribute lfs_migrate script with lustre
133 Details : The lfs_migrate mentioned in the manual is now included in the
134 lustre tree in lustre/scripts and will distributed in lustre RPMs.
136 Severity : enhancement
138 Description: Failover nid list added to import proc file.
140 Severity : enhancement
142 Description: "som_preview" mount option required to use SOM
143 Details : The clients will need to use som_preview mount option
147 Description: Lustre does not do 1MB IOs to HW RAID
148 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
150 Severity : enhancement
152 Description: lfs quota output is not very convenient for awk/sed-parsing
153 Details : Some positions in lfs quota output table could be empty or
154 non-empty which made it hard to parse it with scripts, now
155 a dash is put instead of space where there is not supposed
160 Description: Update to OFED 1.5.1
161 Details : OFED 1.5.1 will be used in environments that don't use
162 vendor supplied OFED.
166 Description: Remove lnet. prefix from lctl params display
167 Details : Also renamed lnet.memused to lnet_memused to
172 Description: Use conf_param -d to remove permanant settings.
176 Description: Add list_param -R
177 Details : lctl list_param now supports recursive mode.
180 Bugzilla : 21829, 22312
181 Description: Fix llobdstat.
182 Details : 1. Now updating the write_bytes counter
183 2. Port script fixes from 1.8.
187 Description: decrease the usage of memory on clients.
188 Details : 1. On clients, recycle dentries and inodes unused.
189 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
194 Description: llobdstat fix and enhancement
195 Details : add a counter to set a limit to how many samples will be returned
196 fix a wildcard in the path to limit to obdfilter stats only
200 Description: lfs check only allowed for root
201 Details : Code cleanups and added a regresion test of non-root use of
206 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
207 Details : Fix hr->hr_index overload.
211 Description: Fix FIEMAP functionality.
215 Description: FC11 patchless client support added
219 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
220 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
224 Description: New acc-sm script and various test-framework enhancments.
228 Description: Previously landed patch was reverted.
232 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
236 Description: Add ldiskfs maxdirsize mount option.
240 Description: Fix mkfs.lustre support for 16TB LUNs.
244 Description: Print mmp_check_interval
245 Details : print mmp_check_interval and make it possible to abort
246 mount operation in case it takes too long.
250 Description: Add cascading_rw to lustre mpi tests.
254 Description: Store SOM data in EA.
255 Details : Also several SOM related cleanups.
259 Description: Add refcount to md_open_data
263 Description: Use refcount on lov_request_set to prevent memory corruption.
267 Description: Statahead: use parent dir's i_mutex for synchronization.
271 Description: loadgen improvements.
275 Description: Enable peer health detection in ksocklnd and ko2iblnd.
279 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
283 Description: Fix race on obd_devcie:md_stats
287 Description: Fix SLES 11 client builds.
288 Details : Resolved uninitalized pointer gcc warnings.
292 Description: Allow non-root access for "lfs check".
293 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
297 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
299 Severity : enhancement
301 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
302 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
306 Description: Sanity test_27p fails.
307 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
308 Sanity tests 27 were cleaned up and improved.
312 Description: Unify req format on client/servers
313 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
314 buffers and swabbers approach..
319 Description: drop join file stuff.
321 Severity : enhancement
323 Description: Add 2.6.27 and SLES11 (patchless client) support
327 Description: Fix in ptlrpc_expire_one_request() to print the signed time
330 Severity : enhancement
332 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
335 Severity : enhancement
337 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
339 Severity : enhancement
341 Description: correctly handle big reply message.
342 Details : send LNet event if reply is bigger then buffer and adjust this buffer
347 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
349 Severity : enhancement
351 Description: Add support for OEL5.
355 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
357 Severity : enhancement
359 Description: "weak-modules" support
360 Details : Implement "weak-modules" support which enables kernel modules
361 to be used with any kernel that implements the same kABI. In
362 order to achieve this modules are now installed in
363 /lib/modules/$(uname -r)/updates/kernel on all distributions.
367 Description: client could not reconnect to OST because of an active request.
368 Details : The client now retries to reconnect to the same server, if a
369 connect request failed with EBUSY or -EAGAIN. Test case 35b
370 conf-sanity.sh has been created.
374 Description: Don't allow make backward step on assignin osc next id.
375 Details : race between allocation next id and ll_sync thread can be cause
376 of set wrong osc next id and can be kill valid ost objects.
380 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
384 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
388 Description: Error handling in osc_statfs_interpret() has been improved.
389 Details : Check in osc_statfs_interpret() for EBADR.
393 Description: Do not send statfs() requests to OSTs disabled by administrator.
394 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
396 Severity : enhancement
398 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
400 Severity : enhancement
402 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
404 Severity : enhancement
406 Description: Enhancement for lfs(1) command to use numeric uid/gid.
410 Description: Improved cleanup procedure for conf-sanity test_46a.
414 Description: Do not start bulk transfer, if client tried to reconnect.
415 Details : Do not start bulk transfer, if client tried to reconnect,
416 but failed due to active rpcs. Check for exp_abort_active_req
417 has been implemented.
421 Description: Prevent inconsistences between linux and lustre mount structures.
422 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
423 Make the sleep interruptible.
425 Severity : enhancement
427 Description: provide server to client comms path
428 Details : server to client communications path via new LDLM_SET_INFO rpc
432 Description: truncate starts GFP_FS allocation under transaction causing deadlock
433 Details : ldiskfs_truncate calls grab_cache_page which may start page
434 allocation under an open transaction. This may lead to
435 calling prune_icache with consequent lustre reentrance.
439 Description: send correctly lsm on open replay
440 Details : MDS is trust to LSM size on replay open, but client can set wrong size
445 Description: enable client interface failover
446 Details : When a child reconnects from another NID, properly update export
447 nid hash position and ldlm reverse import.
449 Severity : enhancement
451 Description: Communicate OST degraded/readonly state via statfs to MDS
452 Details : Flags in the statfs returned from OSTs indicate whether the
453 OST is in a degraded RAID state, or if the filesystem has
454 turned read-only after a filesystem error is detected.
459 Description: don't panic if EPROTO was hit when reading symlink
460 Details : correctly handling request reference in error cases.
462 Severity : enhancement
464 Description: Add LustreNetLink, a kernel-userspace communcation path.
466 Severity : enhancement
468 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
473 Description: access to llog context before init.
474 Details : move handling CATALOGS file at osc layer and forbid access to llog
478 Frequency : rare, only when client is writing to OST on local node
480 Description: Avoid deadlock for local client writes
481 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
482 memory freeing context. This allows OST threads to set the
483 PF_MEMALLOC flag on task structures in order to allocate memory
484 from reserved pools and complete IO.
485 Use GFP_HIGHUSER for OST allocations for non-local client writes,
486 so that the OST threads generate memory pressure and allow
487 inactive pages to be reclaimed.
489 Severity : enhancement
491 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
496 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
497 Details : move ldlm namespace creation in setup phase to avoid grab
498 _lprocfs_lock with cli_sem held.
502 Description: Temporarily disable grant shrink.
503 Details : Disable the feature for debugging.
506 Frequency : only when formatting test filesystems
508 Description: Unable to run several mkfs.lustre on loop devices at the same time
509 Details : mkfs.lustre returns error 256 on the concurrent loop devices
510 formatting. The solution is to proper handle the error.
512 Severity : enhancement
514 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
516 Severity : enhancement
518 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
523 Description: resolve race between obd_disconnect and class_disconnect_exports
524 Details : if obd_disconnect will be called to already disconnected export he
525 forget release one reference and osc module can't unloaded.
527 Severity : enhancement
529 Description: Allow tuning service thread via /proc
530 Details : For each service a new
531 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
532 created that can be used to set min/max thread counts, and get the
533 current number of running threads.
538 Descriptoin: don't return error if have partially created objects for file.
539 Details : lov_update_create_set uses set->set_success as index for created
540 objects, so if some requests will be failed, they will have hole
541 at end of array and qos_shrink_lsm for allocate correct lsm.
543 Severity : enhancement
545 Description: Update OFED support to 1.4.1
550 Description: Reduce small size read RPC
551 Details : Set read-ahead limit for every file and only do read-ahead when
552 available read-ahead pages are bigger than 1M to avoid small size
555 Severity : enhancement
557 Description: move AT tunable parameters for more consistent usage
558 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
560 Severity : enhancement
562 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
563 Details : allow skip disconnected ost for send statfs request and hide error
569 Description: fix racy locking of mballoc block bitmaps causing BUG
570 Details : The locking of the mballoc buddy bitmap and the in-memory
571 block bitmap was using two different spin locks in some
572 cases. This made it possible to incorrectly access the
573 mballoc bitmap while another process was modifying it,
574 causing a sanity assertion to fail. While no on-disk corruption
575 was reported, there was some risk of this happening.
578 Frequency : rare, on llog test 6
580 Descriptoin: don't allow connect to already connected import
581 Details : allowing connect to already connected import is hide connecting problem.
586 Descriptoin: allow kill process which wait statahead result
587 Details : for some reasons 'ls' can stick in waiting result from statahead,
588 in this case need way for kill this process.
590 Severity : enhancement
592 Description: Add state history info file, enhance import info file
593 Details : Track import connection state changes in a new osc/mdc proc file;
594 add overview-type data to the osc/mdc import proc file.
596 Severity : enhancement
598 Description: MDS create should not wait for statfs RPC while holding DLM lock.
600 Severity : enhancement
602 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
607 Descriptoin: OST grant too much space to client even there are not enough space.
608 Details : Client will shrink its grant cache to OST if there are no write
609 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
610 this grant cache if there are already not enough avaible space
611 (left_space < total_clients * 32M).
614 Frequency : start MDS on uncleanly shutdowned MDS device
616 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
617 Details : stay in waiting mds<>ost recovery finished produce random bugs
618 due race between two ll_sync thread for one lov target. send
619 ACTIVATE event only if connect realy finished and import have
623 Frequency : rare, connect and disconnect target at same time
625 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
626 Details : don't call obd_disconnect under lov_lock. this long time
627 operation and can block ptlrpcd which answer to connect request.
632 Descriptoin: don't lose wakeup for imp_recovery_waitq
633 Details : recover_import_no_retry or invalidate_import and import_close can
634 both sleep on imp_recovery_waitq, but we was send only one wakeup
638 Frequency : always with long access acl
640 Descriptoin: mds can't pack reply with long acl.
641 Details : mds don't control size of acl but they limited by reint/getattr
644 Severity : enhancement
646 Description: Update to SLES10 kernel-2.6.16.60-0.33.
648 Severity : enhancement
650 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
653 Frequency : start MDS on uncleanly shutdowned MDS device
655 Descriptoin: aborting recovery hang on MDS
656 Details : don't throttle destroy RPCs for the MDT.
659 Frequency : on remount
661 Description: external journal device not working after the remount
662 Details : clear dev_rdonly flag for external journal devices in
668 Description: shutdown vs evict race
669 Details : client_disconnect_export vs connect request race.
670 if client will evicted at this time - we start invalidate
671 thread without referece to import and import can be freed
675 Frequency : rare, need acl's on inode.
677 Description: client can't handle ost additional correctly
678 Details : if ost was added after client connected to mds client can have
679 hit lnet_try_match_md ... to big messages to wide striped files.
680 in this case need teach client to handle config events about add
681 lov target and update client max ea size at that event.
683 Severity : enhancement
685 Description: Changelogs
686 Details : Changelogs are a lightweight mechanism to track filesystem
687 metadata and namespace changes. The changelog is recorded
688 permanently on the MDTs, and is periodically "consumed" / purged
689 when records are no longer needed.
691 Severity : enhancement
693 Description: compact fld format with extents
694 Details : Store range of seq rather than every seq in FLD. Seq
695 controller update FLD rather than clients. In Case of CMD, mdt0
696 has FLD, all other metadata server act as non persistent proxy
697 for FLD queries and cache fld entries in fld cache.
702 Description: don't skip ost target if they assigned to file
703 Details : Drop slow OSCs if we can, but not for requested start idx.
704 This means "if OSC is slow and it is not the requested
705 start OST, then it can be skipped, otherwise skip it only
706 if it is inactive/recovering/out-of-space.
710 Description: more cleanup in mds_lov
711 Details : not send LOV EA under replay, we can't know about they size at this
712 time. Don't allow client connect to mds before any ost connected,
713 for avoid problems with LOV EA size and returning EIO to client.
715 Severity : enhancement
717 Description: Interoperability at server side (Disk interoperability)
719 Severity : enhancement
721 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
723 Severity : enhancement
725 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
727 Severity : enhancement
729 Description: New client IO stack (CLIO).
731 Severity : enhancement
733 Description: Commit on sharing. Eliminate inter-client dependencies between
734 uncommitted transactions by doing transaction commits.
735 Thereby clients may recovery independently.
738 Frequency : Create a symlink file with a very long name
740 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
741 Details : If there is no extra space in the request for early cancels,
742 ldlm_req_handles_avail() returns 0 instead of a negative value.
744 Severity : enhancement
746 Description: Add /proc entry for import status
747 Details : The mdc, osc, and mgc import directories now have
748 an import directory that contains useful import data for debugging
751 Severity : enhancement
753 Description: Re-disable certain /proc logging
754 Details : Enable and disable client's offset_stats, extents_stats and
755 extents_stats_per_process stats logging on the fly.
758 Frequency : Only on FC kernels 2.6.22+
760 Description: oops in statahead
761 Details : Do not drop reference count for the dentry from VFS when lookup,
762 VFS will do that by itself.
764 Severity : enhancement
766 Description: Generic /proc file permissions
767 Details : Set /Proc file permissions in a more generic way to enable non-
768 root users operate on some /proc files.
772 Description: Hitting mdc_commit_close() ASSERTION
773 Details : Properly handle request reference release in
774 ll_release_openhandle().
778 Description: quota recovery deadlock during mds failover
779 Details : This patch includes att18982, att18236, att18237 in bz14840.
781 1. fix osts hang when mds does failover with quotaon
782 2. prevent watchdog storm when osts threads wait for the
787 Frequency : only patchless client
788 Description: add workaround for race between add/remove dentry from hash
790 Severity : enhancement
792 Description: Allow OST glimpses to return PW locks
796 Description: LBUG when llog conf file is full
797 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
801 Description: Prevent import from entering FULL state when server in recovery
805 Description: service mount cannot take device name with ":"
806 Details : Only when device name contains ":/" will mount treat it as
812 Description: replace ptlrpcd with the statahead thread to interpret the async
813 statahead RPC callback
817 Frequency : on recovery
818 Description: I/O failures after umount during fail back
819 Details : if client reconnected to restarted server we need join to recovery
820 instead of find server handler is changed and process self
821 eviction with cancel all locks.
823 Severity : enhancement
825 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
827 Severity : enhancement
829 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
831 Severity : enhancement
833 Description: Upcall on Lustre log has been dumped
834 Details : Allow for a user mode script to be called once a Lustre log has
835 been dumped. It passes the filename of the dumped log to the
836 script, the location of the script can be specified via
837 /proc/sys/lnet/debug_log_upcall.
842 Description: avoid idr_remove called for id which is not allocated.
843 Details : Move assigment s_dev for clustered nfs to end of initialization,
844 to avoid problem with error handling.
849 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
850 Details : When connection is reused this not moved from CONN_UNUSED_HASH
851 into CONN_USED_HASH and this prodice warning when put connection
852 again in unused hash.
855 Severity : enhancement
857 Description: File striping can now be set to use an arbitrary pool of OSTs.
859 Severity : enhancement
861 Description: Export bytes_read/bytes_write count on OSC/OST.
865 Description: Early reply size mismatch, MGC loses connection
866 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
867 the connect flags are properly negotiated.
872 Description: quotas are not honored with O_DIRECT
873 Details : all writes with the flag O_DIRECT will use grants which leads to
874 this problem. Now using OBD_BRW_SYNC to guard this.
878 Description: add quota statistics
879 Details : 1. sort out quota proc entries and proc code.
880 2. add quota statistics
882 Severity : enhancement
884 Description: enable quota support for HEAD.
888 Description: Properly propagate oinfo flags from lov to osc for statfs
889 Details : restore missing copy oi_flags to lov requests.
891 Severity : enhancement
893 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
894 : l_getgroups(8), lst(8), routerstat(8)
895 Details : included man pages for llobdstat(8), llstat(8),
896 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
898 Severity : enhancement
900 Description: configure's --enable-quota should check the
901 : kernel .config for CONFIG_QUOTA
902 Details : configure is terminated if --enable-quota is passed but
903 : no quota support is in kernel
907 Description: Remove portals compatibility
908 Details : Remove portals compatibility, not interoperable with releases
913 Description: Resolve device initialization race
914 Details : Prevent proc handler from accessing devices added to the
915 obd_devs array but yet be intialized.
917 Severity : enhancement
919 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
921 Severity : enhancement
923 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
928 Description: Using wrong pointer in osc_brw_prep_request
929 Details : Access to array[-1] can produce panic if kernel compiled with
930 CONFIG_PAGE_ALLOC enabled
934 Description: Client runs out of low memory
935 Details : Consider only lowmem when counting initial number of llap pages
939 Description: *optional* service tags registration
940 Details : if the "service tags" package is installed on a Lustre node
941 When the filesystem is mounted, a local-node service tag will
942 be created. See http://inventory.sun.com/ for more information
943 about the Service Tags asset management system.
947 Description: Kernel BUG tries to release flock
948 Details : Lustre does not destroy flock lock before last reference goes
949 away. So always drop flock locks when client is evicted and
950 perform unlock regardless of successfulness of speaking to MDS.
954 Description: add refcount protection for osc callbacks, avoid panic on shutdown
958 Description: sanity test 65a fails if stripecount of -1 is set
959 Details : handle -1 striping on filesystem in ll_dirstripe_verify
964 Description: ASSERTION(CheckWriteback(page,cmd)) failed
965 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
968 Severity : enhancement
970 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
974 Description: do not process already freed flock
975 Details : flock can possibly be freed by another thread before it reaches
976 to ldlm_flock_completion_ast.
980 Description: LBUG during stress test
981 Details : Need properly lock accesses the flock deadlock detection list.
985 Description: oops in page fault handler
986 Details : kernel page fault handler can return two special 'pages' in error
987 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
991 Description: timeout with invalidate import.
992 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
993 should be handled by ptlrpcd. This produce long age waiting and
994 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
996 Severity : enhancement
998 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1002 Description: enable MGS and MDT services start separately
1003 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1004 a co-located MGS without starting the MGS, which is a complement
1005 to 'nosvc' mount option.
1009 Frequency : after recovery
1010 Description: precreate to many object's after del orphan.
1011 Details : del orphan st in oscc last_id == next_id and this triger growing
1012 count of precreated objects. Set flag LOW to skip increase count
1013 of precreated objects.
1017 Frequency : rare, on clear nid stats
1018 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1019 Details : when clean nid stats sometimes try destroy live entry,
1020 and this produce panic in free.
1024 Description: Stack overflow during MDS log replay
1025 ease stack pressure by using a thread dealing llog_process.
1029 Description: wait until IO finished before start new when do lock cancel.
1030 Details : VM protocol want old IO finished before start new, in this case
1031 need wait until PG_writeback is cleared until check dirty flag and
1032 call writepages in lock cancel callback.
1034 Severity : enhancement
1036 Description: using special macro for print time and cleanup in includes.
1040 Description: mds_mfd_close() ASSERTION(rc == 0)
1041 Details : In mds_mfd_close(), we need protect inode's writecount change
1042 within its orphan write semaphore to prevent possible races.
1046 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1050 Frequency : rare, on shutdown ost
1051 Description: don't hit live lock with umount ost.
1052 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1053 use shrink_dcache_sb instead.
1057 Description: don't panic with use echo client
1058 Details : echo client pass NULL as client nid pointer and this produce null
1059 pointer dereference.
1063 Description: fix build on ppc32
1064 Details : compile code with -m64 flag produce wrong object file for ppc32.
1068 Description: add message levels for liblustreapi
1072 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1073 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1074 will result in an -EAGAIN and in the exit call path we call
1075 alloc_rr() which will with increasing aggressiveness attempt to
1076 aquire precreated objects on the minimum number of required OSCs.
1080 Description: Use old size assignment to avoid deadlock
1081 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1082 the scheduling while holding a spinlock. We do not need locking
1083 for size in ll_update_inode() because size is only updated from
1084 the MDS for directories or files without objects, so there is no
1085 other place to do the update, and concurrent access to such inodes
1086 are protected by the inode lock.
1090 Description: resolve "_IOWR redefined" build error on SLES10
1094 Description: dump the memory debugging after all modules are unloaded to
1095 suppress false negative in conf_sanity test 39
1097 Severity : enhancement
1099 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1104 Description: fix SLES kernel versioning
1105 Details : the kernel version for our SLES 10 kernel did not include a "-"
1106 before the "smp" at the end. while this was not a problem in
1107 general, it did mean that software trying to use the kernel
1108 version to try to detect a vendor specific kernel would fail.
1109 this was most evident by the OFED build scripts.
1113 Description: Don't update lov_desc members until making sure they are valid
1114 Details : When updating lov_desc members via proc fs, need fix their
1115 validities before doing the real update.
1119 Description: don't put request into delay list while invalidate in flight.
1120 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1121 import in flight. this produce timeout for invalidate and sometimes
1122 can cause stale data.
1126 Frequency : on ppc only
1127 Description: not convert ost objects for directory because it's not exist.
1128 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1132 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1134 Severity : enhancement
1136 Description: Update to support 2.6.22.14 vanilla kernel.
1140 Frequency : rare, on recovery
1141 Description: read procfs can produce deadlock in some situation
1142 Details : Holding lprocfs lock which send rpc can produce block for destroy
1143 obd objects and this also block reconnect with -EALREADY. This
1144 isn't fix all lprocfs bugs - but make it rare.
1146 Severity : enhancement
1148 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1151 Frequency : frequent on X2 node
1153 Description: mdc_set_open_replay_data LBUG
1154 Details : Set replay data for requests that are eligible for replay.
1158 Description: lustre_mgs: operation 101 on unconnected MGS
1159 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1160 MGC, and late on MGC cannot successfully connect to MGS and a lot
1161 of the error messages complaining that MGS is not connected.
1165 Frequency : on network error
1166 Description: panic with double free request if network error
1167 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1168 but it's true only for synchronus enqueue, for async enqueue
1169 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1172 Severity : enhancement
1174 Description: client-side metadata stat-ahead during readdir(directory readahead)
1175 Details : perform client-side metadata stat-ahead when the client detects
1176 readdir and sequential stat of dir entries therein
1179 Frequency : on start mds
1181 Description: Implement get_info(last_id) in obdfilter.
1184 Frequency : occasional
1186 Description: Correctly check stale fid, not start epoch if ost not support SOM
1187 Details : open with flag O_CREATE need set old fid in op_fid3 because
1188 op_fid2 was overwritten with new generated fid, but mds can answer
1189 with one of these two fids and both is not stale. Setattr
1190 incorrectly started an epoch and assume will be called
1191 done_writing, but without SOM done_writing ever being called.
1194 Frequency : rare, depends on device drivers and load
1196 Description: MDS or OSS nodes crash due to stack overflow
1197 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1198 In some cases, in conjunction with device drivers that use a lot
1199 of stack the MDS (or possibly OSS) service threads could overflow
1200 the stack. One change which was identified to consume additional
1201 stack has been reworked to avoid the extra stack usage.
1204 Frequency : occasional
1206 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1207 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1208 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1209 oscc_has_objects and see OSC has no objests and call
1210 oscc_internal_create to resend create request.
1212 Severity : enhancement
1214 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1216 Severity : enhancement
1218 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1221 Frequency : very rare
1223 Description: Fix replay if there is an un-replied request and open
1224 Details : In some cases, older replay request will revert the
1225 mcd->mcd_last_xid on MDS which is used to record the client's
1226 latest sent request.
1228 Severity : enhancement
1230 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1232 Severity : enhancement
1234 Description: Add rhel5 support to HEAD.
1236 Serverity : enhancement
1238 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1243 Description: Don't allow skipping OSTs if index has been specified.
1244 Details : Don't allow skipping OSTs if index has been specified, make
1245 locking in internal create lots better.
1249 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1250 Details : access to bitfield in structure is always rounded to long
1251 and this produce problem with not atomic change any bit.
1255 Description: Lustre make rpms failed.
1256 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1257 builds Lustre rpms from tarball.
1260 Frequency : rare on shutdown ost
1262 Description: If llog cancel was not send before clean_exports phase, this can
1263 produce deadlock in llog code.
1264 Details : If llog thread has last reference to obd and call class_import_put
1265 this produce deadlock because llog_cleanup_commit_master wait when
1266 last llog_commit_thread exited, but this never success because was
1267 called from llog_commit_thread.
1271 Description: allow userland application know is lost one of stripes.
1272 Details : fill lvb_blocks with error code on ost and return it to
1273 application if error flag found.
1277 Description: NULL lov_tgts causing MDS oops
1278 Details : more safe checks for NULL lov_tgts for avoid oops.
1280 Severity : enhancement
1282 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1286 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1287 Details : If an OST is down the MDS will hang indefinitely in
1288 obd_statfs_rqset() waiting for the statfs data. While for
1289 MDS QOS usage of statfs, it should not stuck in waiting.
1291 Severity : enhancement
1293 Description: remote_acl support
1294 Details : Support ACL-based permission check for remote user.
1295 Support setfacl/getfacl for remote user with the utils
1296 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1297 the system "{s,g}etfacl" utils.
1299 Severity : enhancement
1301 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1303 Severity : enhancement
1305 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1309 Description: make mgs_setparam() handle fsname containing dash
1310 Details : fsname containing a dash does not work with lctl conf_param
1312 Severity : enhancement
1314 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1316 Severity : enhancement
1318 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1322 Frequency : rare, at shutdown
1323 Description: access already free / zero obd_namespace.
1324 Details : if client_disconnect_export was called without force flag set,
1325 and exist connect request in flight, this can produce access to
1326 NULL pointer (or already free pointer) when connect_interpret
1327 store ocd flags in obd_namespace.
1331 Frequency : only at startup
1332 Description: not alloc memory with spinlock held.
1333 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1334 if any spinlock held.
1336 Severity : enhancement
1338 Description: make lustre randomly fail allocating memory
1339 Details : Make lustre randomly failed allocating memory for testing purpose.
1341 Severity : enhancement
1343 Description: lost problems with lov objid file
1344 Details : Fixes some scability and access to not inited memory problems
1345 in work with lov objdid file.
1350 Description: lfs find does not continue on file error
1351 Details : Continue other files processing when a file/dir is absent.
1355 Description: Inconsistent usage of lustre_pack_reply()
1356 Details : Standardize the usage of lustre_pack_reply() such that it
1357 always generate a CERROR on failure.
1360 Frequency : occasional
1362 Description: MDS hang or stay in waiting lock
1363 Details : If client receive lock with CBPENDING flag ldlm need send lock
1364 cancel as separate rpc, to avoid situation when cancel request
1365 can't processed due all i/o threads stay in wait lock.
1369 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1373 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1377 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1381 Description: Update to RHEL4 latest kernel.
1383 Severity : enhancement
1385 Description: Build SLES10 patchless client fails
1386 Details : The configure was broken by run ./configure with
1387 --with-linux-obj=.... argument for patchless client. When the
1388 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1389 can't search header adequately. Use absolute path such as
1390 -I($LINUX)/include instead.
1394 Description: interrupt oig_wait produce painc on resend.
1395 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1396 private set, and this produce situation when rq_ptlrpcd_data not
1397 copyed to new allocated request and triggered LBUG on assert
1398 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1399 ptlrpcd set if request is changed and can be safety changed to use
1404 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1405 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1406 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1407 these macros and replace them with their difinitions instead.
1411 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1412 Details : Update which_patch & target file for SLES10 latest kernel.
1415 Description: Remove client patches from SLES 10 kernel.
1416 Details : This causes SLES 10 clients to behave as patchless clients
1417 even on a Lustre-patched (server) kernel.
1419 Severity : enhancement
1421 Description: self-adjustable client's lru lists
1422 Details : use adaptive algorithm for managing client cached locks lru
1423 lists according to current server load, other client's work
1424 pattern, memory activities, etc. Both, server and client
1425 side namespaces provide number of proc tunables for controlling
1428 Severity : enhancement
1430 Description: light-weight GSS support
1431 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1432 and reduce performance overhead.
1434 Severity : enhancement
1436 Description: Linux keyring support
1437 Details : Support using service of Linux keyring for Lustre GSS internal
1438 context refresh/cache mechanism.
1442 Description: Fix errors in lfs documentation
1443 Details : Fixes man pages
1447 Description: don't use GFP_* in generic Lustre code.
1448 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1452 Description: obdclass is limited by single OBD_ALLOC(idarray)
1453 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1457 Description: warnings with build patchless client with vanila 2.6.19 and up
1458 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1459 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1463 Frequency : Only for SLES
1464 Description: sanity_quota fail test_1
1465 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1466 file, which makes TSTID[2] inunique.
1470 Frequency : Only for SLES9
1471 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1472 Details : Update target/ChangeLog/which_patch .
1476 Frequency : Only for test_10 in sanity-quota.sh
1477 Description: a bug in quota test code
1478 Details : checking a test flag in a wrong place
1482 Frequency : Only for test_13 in sanity-quota.sh
1483 Description: a bug in quota test script
1484 Details : shouldn't check overfull allocated quota
1488 Description: jbd statistics
1489 Details : Port older jbd statistics patch for sles10
1491 Severity : enhancement
1493 Description: Add printing inode info into message about error in writepage.
1498 Description: reply_lock_interpret crash due to race with it and lock cancel.
1499 Details : Do not replay locks that are being cancelled. Do not reference
1500 locks by their address during replay, just by their handle.
1504 Frequency : When flocks are used.
1505 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1506 Details : Flock locks might destroy just granted lock if it could be merged
1507 with another existing flock, this is done in completion handler,
1508 so teach ldlm_cli_enquque_fini that this is a valid case for
1514 Description: Oops in read and write path when failing to allocate lock.
1515 Details : Check if lock allocation failed and return error back.
1519 Description: lstripe command fails for valid OST index
1520 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1521 instead of lov->desc.ld_active_tgt_count.
1525 Description: sanity.sh failed test 103
1526 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1531 Description: df doesn't work properly if diskfs blocksize != 4K
1532 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1536 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1537 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1539 Severity : enhancement
1541 Description: Adaptive timeouts
1542 Details : RPC timeouts adapt to changing server load and network
1543 conditions to reduce resend attempts and improve recovery time.
1547 Description: llapi_file_create() does not allow some changes
1548 Details : add llapi_file_open() that allows specifying the mode and
1549 open flags, and also returns an open file handle.
1553 Description: merge and cleanup kernel patches.
1554 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1558 Description: Add journal checksum support.(Kernel part)
1559 Details : The journal checksum feature adds two new flags i.e
1560 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1561 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1562 indicates that the commit block contains the checksum for
1563 the blocks described by the descriptor blocks. Now commit
1564 record can be sent to disk without waiting for descriptor
1565 blocks to be written to disk. This behavior is controlled
1566 using JBD2_FEATURE_ASYNC_COMMIT flag.
1570 Description: OSS needs mutliple precreate threads
1571 Details : Add ability to start more than one create thread per OSS.
1575 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1576 Details : execute lfs setstripe on client
1580 Description: mds_obd_create error creating tmp object
1581 Details : When the user sets quota on root, llog will be affected and can't
1582 create files and write files.
1585 Frequency : Always on ia64 patchless client, and possibly others.
1587 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1588 Details : This allows the patchless client to be loaded on architectures
1589 without this export.
1593 Description: RedHat Update kernel for RHEL5
1594 Details : Add the kernel config file for RHEL5.
1598 Description: RedHat Update kernel for RHEL5
1599 Details : Modify the kernel config file more closer RHEL5.
1603 Description: Build failure against Centos5 (RHEL5)
1604 Details : Define PAGE_SIZE when it isn't present.
1608 Description: test_7 on sanity-quota.sh can't be used on separate machines
1609 Details : add facet to handle it
1613 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1614 Details : with NFS, the anon dentry's parent was set to itself in
1615 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1616 corresponding dentry other than use rec->ur_name.
1618 Severity : enhancement
1620 Description: lfs setstripe enhancement
1621 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1625 Description: enable data checksumming by default
1626 Details : enable checksum by default, allow --disable-checksum
1627 configure option and "-o nochecksum" mount option. Checksums
1628 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1632 Description: lustre support for RHEL5
1633 Details : Add support for RHEL5.
1637 Description: Client eviction due to failover config
1638 Details : after a connection loss, the lustre client should attempt to
1639 reconnect to the last active server first before trying the
1640 other potential connections.
1644 Description: when mds and osts use different quota unit(32bit and 64bit),
1645 quota will be released repeatly.
1646 Details : void sending multiple quota reqs to mds, which will keep the
1647 status between the reqs.
1651 Description: osts not allocated evenly to files
1652 Details : change the condition to increase offset_idx
1655 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1657 Description: Data corruption for OSTs that are formatted larger than 2TB
1659 Details : When generating the bio request for lustre file writes the
1660 sector number would overflow a temporary variable before being
1661 used for the IO. The data reads correctly from Lustre (which
1662 will overflow in a similar manner) but other file data or
1663 filesystem metadata may be corrupted in some cases.
1667 Description: Tune the kernel for good SCSI performance.
1668 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1669 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1674 Description: rewrite ext2-derived code in obdclass/uuid.c
1675 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1676 from scratch preserving functionality.
1680 Description: "lfs find -obd UUID" prints directories
1681 Details : "lfs find -obd UUID" will return all directory names instead
1682 of just file names. It is incorrect because the directories
1683 do not reside on the OSTs.
1687 Description: use i_size_read and i_size_write in 2.6 port
1688 Details : replace inode->i_size access with i_size_read/write()
1691 Frequency : when removing large files
1693 Description: scheduling issue during removal of large Lustre files
1694 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1695 It causes scheduling issues when removing large files (17TB in the
1699 Frequency : only with liblustre clients on XT3
1701 Description: evictions taking too long
1702 Details : allow llrd to evict clients directly on OSTs
1705 Frequency : only on ppc
1707 Description: /proc/fs/lustre/devices broken on ppc
1708 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1709 We should make sure the type of 'index' is loff_t and then cast
1710 explicitly as needed below. Do not assign an explicitly cast
1714 Frequency : only for rhel5
1716 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1717 Details : Modify the target file & which_kernel.
1719 Severity : enhancement
1721 Description: omit set fsid for export NFS
1722 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1723 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1727 Frequency : Always, for kernels after 2.6.16
1728 Description: Fix warning idr_remove called for id=.. which is not allocated.
1729 Details : Last kernels save old s_dev before kill super and not allow
1730 to restore from callback - restore it before call kill_anon_super.
1734 Description: buffer overruns could theoretically occur
1735 Details : llapi_semantic_traverse() modifies the "path" argument by
1736 appending values to the end of the origin string, and a
1737 overrun may occur. Adding buffer overrun check in liblustreapi.
1741 Description: Fix error on 'ls .' at the top of the Lustre mount.
1742 Details : Don't revalidate dentry if it is a root dentry.
1746 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1747 Details : Modify vm-tunables-rhel4.patch.
1751 Description: Kernel config for 2.6.18-vanilla.
1752 Details : Modify targets/2.6-vanilla.target.in.
1753 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1754 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1755 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1756 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1760 Description: improve handling recoverable errors
1761 Details : if request processig with error which can be recoverable on server
1762 request should be resend, otherwise page released from cache and
1767 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1768 Details : Modify target file & which_patch.
1769 A flaw was found in the IA32 system call emulation provided
1770 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1771 value could be stored in the %RAX register, which could trigger an
1772 out-of-bounds system call table access. An untrusted local user
1773 could exploit this flaw to run code in the kernel
1774 (ie a root privilege escalation). (CVE-2007-4573).
1778 Description: O_DIRECT bypasses client statistics.
1779 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1780 read_ahead_stats were updated. copied the stats section from
1781 osc_send_oap_rpc() into async_internal().
1785 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1789 Description: change order of libsysio includes
1790 Details : '#include sysio.h' should always come before '#include xtio.h'
1794 Description: handle "serious error: objid * already exists" more gracefully
1795 Details : If LAST_ID value on disk is smaller than the objects existing in
1796 the O/0/d* directories, it indicates disk corruption and causes an
1797 LBUG(). If the object is 0-length, then we should use the existing
1798 object. This will help to avoid a full fsck in most cases.
1800 Severity : enhancement
1802 Description: adapt the lustre_config script to support the upgrade case
1803 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1804 server targets to 1.6 in parallel.
1808 Description: To avoid grant space > avaible space when the disk is almost
1809 full. Without this patch you might see the error "grant XXXX >
1810 available" or some LBUG about grant, when the disk is almost
1812 Details : In filter_check_grant, for non_grant cache write, we should
1813 check the left space by if (*left > ungranted + bytes), instead
1814 of (*left > ungranted), because only we are sure the left space
1815 is enough for another "bytes", then the ungrant space should be
1816 increase. In client, we should update cl_avail_grant only there
1817 is OBD_MD_FLGRANT in the reply.
1821 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1822 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1823 described in CVE-2007-4573. This problem could allow untrusted
1824 local users to gain root access.
1827 Frequency : when using O_DIRECT and quotas
1829 Description: Incorrect file ownership on O_DIRECT output files
1830 Details : block usage reported by 'lfs quota' does not take into account
1831 files that have been written with O_DIRECT.
1836 Description: touch file failed when fs is not full
1837 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1838 otherwise we can get ENOSP while fs is not full.
1842 Description: parallel lock callbacks
1843 Details : Instead of sending blocking and completion callbacks as separated
1844 requests, adding them to a set and sending in parallel.
1847 Frequency : only for Cray XT3
1848 Bugzilla : 12829/13455
1849 Description: Changing primary group doesn't change the group lustre assigns to
1851 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1852 overridden with the primary group provided by the group upcall,
1853 whereas the supplied fsgid can be trusted if it is in the list of
1854 supplementary groups returned by the group upcall.
1856 Severity : enhancement
1858 Description: Allow masking D_WARNING, D_ERROR messages from console
1859 Details : Console messages can now be disabled via lnet.printk.
1863 Description: User code with malformed file open parameter crashes client node
1864 Details : Before packing join_file req, all the related reference should be
1865 checked carefully in case some malformed flags cause fake
1866 join_file req on client.
1870 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1871 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1872 mode is equal to the granted mode. However, it should also check
1873 that LDLM_FL_CP_REQD is not set.
1877 Description: Heavy nfs access might result in deadlocks
1878 Details : After ELC code landed, it is now improper to enqueue any mds
1879 locks under och_sem, because enqueue might want to decide to
1880 cancel open locks for same inode we are holding och_sem for.
1884 Description: Client eviction while running blogbench
1885 Details : A lot of unlink operations with concurrent I/O can lead to a
1886 deadlock causing evictions. To address the problem, the number of
1887 oustanding OST_DESTROY requests is now throttled to
1888 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1889 callbacks are processed in priority.
1893 Description: enable ACLs on MDS by default
1894 Details : ACLs must be enabled on MDS by default.
1897 Frequency : PPC/PPC64 only
1899 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1900 Details : fix duplicated definitions between asm-ppc64/types.h and
1901 lustre_types.h on PPC.
1904 Frequency : PPC/PPC64 only
1906 Description: asm-ppc/segment.h does not exist
1907 Details : fix compile issue on PPC.
1911 Description: better handle error messages in extents code
1914 Frequency : RHEL4 only
1916 Description: mkfs is very slow on IA64/RHEL4
1917 Details : A performance regression has been discovered in the MPT Fusion
1918 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1919 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1920 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1922 Severity : enhancement
1924 Description: SNMP support enhancement
1925 Details : Adding total number of sampled request for an MDS node in snmp
1928 Severity : enhancement
1930 Description: Optimize ldlm waiting list processing for PR extent locks
1931 Details : When processing waiting list for read extent lock and meeting read
1932 lock that is same or wider to it that is not contended, skip
1933 processing rest of the list and immediatelly return current
1934 status of conflictness, since we are guaranteed there are no
1935 conflicting locks in the rest of the list.
1939 Description: Time out and refuse to reconnect
1940 Details : When the failover node is the primary node, it is possible
1941 to have two identical connections in imp_conn_list. We must
1942 compare not conn's pointers but NIDs, otherwise we can defeat
1943 connection throttling.
1947 Description: port llog fixes from b1_6 into HEAD
1948 Details : Port llog reference couting and some llog cleanups from b1_6
1949 (bug 10800) into HEAD, for protect from panic and access to already
1950 free llog structures.
1954 Description: Detect stride IO mode in read-ahead
1955 Details : When a client does stride read, read-ahead should detect that and
1956 read-ahead pages according to the detected stride pattern.
1960 Description: data checksumming impacts single node performance
1961 Details : add support for several checksum algorithm. Currently, only CRC32
1962 and Adler-32 are supported. The checksum type can be changed on
1963 the fly via /proc/fs/lustre/osc/*/checksum_type.
1967 Description: use adler32 for page checksums
1968 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1973 Description: build for x2 fails
1974 Details : fix compile issue on Cray systems.
1978 Description: Properly match for duplicate locks
1979 Details : Due to different lock order from skiplists code, we need to
1980 traverse entire list for now
1983 Frequency : only on PPC/SLES10
1985 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1986 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1987 be defined. Add a hack in mkfs_lustre.c to work around this compile
1992 Description: LASSERT on MDS when client holding flock lock dies
1993 Details : ldlm pool logic depends on number of granted locks equal to
1994 number of released locks which is not true for flock locks, so
1995 just exclude such locks from consideration.
1999 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2000 Details : Use fsfilt_sync() for both the whole filesystem sync and
2001 individual file sync to eliminate dangerous inode locking
2002 with I_LOCK that can lead to a deadlock.
2006 Description: performance in 1.6.3
2007 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2008 sure that 1MB requests are not fragmented by the block layer.
2010 Severity : enhancement
2012 Description: organize the server-side client stats on per-nid basis
2013 Details : Change the structure of stats under obdfilter and mds to
2021 The "uuid"s file would list the uuids of _active_ exports.
2022 And the clear entry is to clear all stats and stale nids.
2024 Severity : enhancement
2026 Description: eliminate client locks in face of contention
2027 Details : file contention detection and lockless i/o implementation
2028 for contended files.
2032 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2035 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2037 Description: MDS may lose file striping (and hence file data) in some cases
2038 Details : If there are additional extended attributes stored on the MDS,
2039 in particular ACLs, SELinux, or user attributes (if user_xattr
2040 is specified for the client mount options) then there is a risk
2041 of attribute loss. Additionally, the Lustre file striping
2042 needs to be larger than default (e.g. striped over all OSTs),
2043 and an additional attribute must be stored initially in the
2044 inode and then increase in size enough to be moved to the
2045 external attribute block (e.g. ACL growing in size) for file
2050 Description: skiplist implementation simplification
2051 Details : skiplists are used to group compatible locks on granted list
2052 that was implemented as tracking first and last lock of each
2053 lock group the patch changes that to using doubly linked lists
2057 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2058 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2059 because LASSERT on both the data supplied by a client, and the
2060 data on disk is dangerous and incorrect.
2062 Severity : enhancement
2064 Description: Slow truncate/writes to huge files at high offsets.
2065 Details : Directly associate cached pages to lock that protect those pages,
2066 this allows us to quickly find what pages to write and remove
2067 once lock callback is received.
2071 Description: more ldlm soft lockups
2072 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2073 starve other threads from the resource lock for a long time in
2074 case of long waiting queue, so change the debug level from
2075 D_OTHER to the less frequently used D_INFO.
2077 Severity : enhancement
2079 Description: add -gid, -group, -uid, -user options to lfs find
2083 Description: Hung threads in invalidate_inode_pages2_range
2084 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2085 once one is completed. As a result, some RPCs are stuck in the
2086 queue and are never sent.
2090 Description: filter threads hungs on waiting journal commit
2091 Details : Cleanup filter group llog code, then only filter group llog will
2092 be only created in the MDS/OST syncing process.
2096 Description: Procfs and llog threads access destoryed import sometimes.
2097 Details : Sync the import destoryed process with procfs and llog threads by
2098 the import refcount and semaphore.
2100 Severity : enhancement
2102 Description: openlock cache of b1_6 port to HEAD
2107 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2108 Details : If insertion of an extent fails, then discard the inode
2109 preallocation and free data blocks else it can lead to duplicate
2114 Description: don't always update ctime in ext3_xattr_set_handle()
2115 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2116 In some cases the ctime should not be updated, for example for
2117 2.0->1.8 compatibility it is necessary to delete an xattr and it
2118 should not update the ctime.
2122 Bugzilla : 15713/16362
2123 Description: Assertion in iopen_connect_dentry in 1.6.3
2124 Details : looking up an inode via iopen with the wrong generation number can
2125 populate the dcache with a disconneced dentry while the inode
2126 number is in the process of being reallocated. This causes an
2127 assertion failure in iopen since the inode's dentry list contains
2128 both a connected and disconnected dentry.
2132 Description: assertion failure in ldlm_handle2lock()
2133 Details : fix a race between class_handle_unhash() and class_handle2object()
2134 introduced in lustre 1.6.5 by bug 13622.
2139 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2140 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2145 Description: oops in sd_iostats_seq_show()
2146 Details : unloading/reloading the scsi low level driver triggers a kernel
2147 bug when trying to access the sd iostat file.
2152 Description: Kernel panics during QLogic driver reload
2153 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2154 patch, causing memory corruption.
2159 Description: journal_dev option does not work in b1_6
2160 Details : pass mount option during pre-mount.
2162 Severity : enhancement
2164 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2165 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2166 extent information of a file. It can be used to map logical blocks
2167 in a file to physical blocks in the block device.
2171 Description: LDLM soft lockups - improvement
2172 Details : It is be possible to send the lock handle along with each read
2173 or write request because the client is already doing a lock match
2174 itself so there isn't any reason the OST should have to re-do that
2178 Frequency : only on Cray X2
2180 Description: X2 build failures
2181 Details : fix build failures on Cray X2.
2185 Description: xid & resent requests
2186 Details : Initialize RPC XID from clock at startup (randomly if clock is
2189 Severity : enhancement
2191 Description: Add lustre_start utility to start or stop multiple Lustre servers
2196 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2197 Details : In case of memory pressure, list_del() can be called twice on
2198 req->rq_history_list, causing a kernel oops.
2202 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2203 Details : fix stack overflow in the distributed lock manager by defering
2204 export eviction after a failed AST to the elt thread instead of
2205 handling it in the dlm interpret routine.
2209 Description: Convert some comments to new format.
2210 Details : Update documenting comments to match doxygen conventions.
2214 Description: Grammar fixes.
2215 Details : A couple of trivial spelling fixes.
2219 Description: OSD_COUNTERS-mandatory
2220 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2221 to simplify the code.
2225 Description: simplify cmm_device freeing logic.
2226 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2230 Description: Add lockdep support to dt_object_operations locking interface.
2231 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2232 parameter indicating lock ordering. Update mdd code to use new
2237 Description: Introduce failloc constants for lockless IO tests.
2238 Details : Add two new failloc constants to test lockless IO. Only one of
2239 them in implemented---another is checked in yet to be landed
2244 Description: Add lockdep support for inode mutex.
2245 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2246 in the situations where Lustre has to lock more than one inode
2251 Description: Add optional invariants checking support.
2252 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2253 configure switch. This macro is to be used for consistency and
2254 sanity checks that are too expensive to be left in `production'
2259 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2260 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2265 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2266 Details : Use spin_lock_nested() in (the only) situation where more than
2267 one ldlm_lock is locked simultaneously. Also, fix possible
2268 dead-lock in ldlm_lock_change_resource() by enforcing particular
2273 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2274 Details : Instead of specifying each ldlm_lock call-back through separate
2275 parameter, wrap them into struct ldlm_callback_suite.
2279 Description: Kill join_lru obd method and its callers.
2280 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2281 in memory---a requirement implemented through ->o_join_lru() obd
2282 method in HEAD. Remove this method and its users.
2286 Description: Add asynchronous ldlm ENQUEUE completion handler.
2287 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2288 case that stock ldlm_completion_ast() cannot handle as it waits
2289 until lock is granted. Introduce new ldlm_completion_ast_async()
2290 for this. Also comment ldlm_completion_ast().
2294 Description: ldlm_error <-> errno conversion.
2295 Details : Add functions to map (rather arbitrary) between LDLM error codes
2296 and standard errno values. CLIO needs this to prevent LDLM specific
2297 constants from escaping ldlm and osc.
2301 Description: Kill unused ldlm_handle2lock_ns() function.
2302 Details : Kill unused ldlm_handle2lock_ns() function.
2306 Description: Add lu_ref support to ldlm_lock
2307 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2308 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2309 and ldlm_resource. LDLM interface has to be changed, because code that
2310 releases a reference on a lock, has to "know" what reference this is.
2311 In the most frequent case
2313 lock = ldlm_handle2lock(handle);
2315 LDLM_LOCK_PUT(lock);
2317 no changes are required. When any other reference (received _not_ from
2318 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2319 instead of LDLM_LOCK_PUT().
2321 Arguably, changes are pervasive, and interface requires some discipline
2322 for proper use. On the other hand, it was very instrumental in finding
2323 a few leaked lock references.
2327 Description: Add ldlm_lock_addref_try().
2328 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2329 attempts to addref a lock that might be being canceled
2334 Description: Add ldlm_weigh_callback().
2335 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2336 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2337 hard-coded `number of pages' logic.
2341 Description: Add lockdep annotations to llog code.
2342 Details : Use appropriately tagged _nested() locking calls in the places
2343 where llog takes more than one ->lgh_lock lock.
2347 Description: Add loi_kms_set().
2348 Details : Wrap kms updates into a helper function.
2352 Description: Constify instances of struct lsm_operations.
2353 Details : Constify instances of struct lsm_operations.
2357 Description: lu_conf support.
2358 Details : On a server, a file system object is uniquely identified
2359 by a fid, which is sufficient to locate and load all object
2360 state (inode). On a client, on the other hand, more data are
2361 necessary instantiate an object. Change lu_object_find() and
2362 friends to take additional `lu_conf' argument describing object.
2363 Typically this includes layout information.
2367 Description: lu_context fixes.
2368 Details : Introduce new lu_context functions that are needed on the client
2369 side, where some system threads (ptlrpcd) are shared by multiple
2370 modules, and so cannot be stopped during module shutdown.
2374 Description: Add start and stop methods to lu_device_type_operations.
2375 Details : Introduce two new methods in lu_device_type_operations, that are
2376 invoked when first instance of a given type is created and last one
2377 is destroyed respectively. This is need by CLIO.
2381 Description: Add lu_ref support to struct lu_device.
2382 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2383 track leaked references.
2387 Description: Introduce lu_kmem_descr.
2388 Details : lu_kmem_descr and its companion interface allow to create
2389 and destroy a number of kmem caches at once.
2393 Description: Fix lu_object finalization race.
2394 Details : Fix a race between lu_object_find() finding an object and its
2395 concurrent finalization. This race is (most likely) not possible
2396 on the server, but might happen on the client.
2400 Description: Introduce lu_ref interface.
2401 Details : lu_ref is a debugging module allowing to track references to
2402 a given object. It is quite cpu expensive, and has to be
2403 explicitly enabled with --enable-lu_ref. See usage description
2408 Description: Factor lu_site procfs stats into a separate function.
2409 Details : Separate lu_site stats printing code into a separate function
2410 that can be reused on a client.
2414 Description: Constify instances of struct {lu,dt,md}_device_operations.
2415 Details : Constify instances of struct {lu,dt,md}_device_operations.
2419 Description: Introduce struct md_site and move meta-data specific parts of
2420 struct lu_site here.
2421 Details : Move md-specific fields out of struct lu_site into special struct
2422 md_site, so that lu_site can be used on a client.
2426 Description: Kill mdd_lov_destroy().
2427 Details : Remove unused mdd code.
2431 Description: Add st_block checking to multistat.c.
2432 Details : Add st_block checking to multistat.c.
2436 Description: Add lu_ref support to struct obd_device.
2437 Details : Add lu_ref tracking to obd_device.
2441 Description: Kill obd_set_fail_loc().
2442 Details : Remove unused code.
2446 Description: Add special type for ptlrpc_request interpret functions.
2447 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2448 there. Actual usage will be in CLIO.
2452 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2453 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2454 doesn't work with lockdep.
2458 Description: Add rwv.c test program.
2459 Details : New testing program exercising readv(2) and writev(2) (Qian).
2463 Description: Add sendfile.c test program.
2464 Details : New testing program exercising sendfile(2) (Jay).
2468 Description: Ratelimit a message that can be very frequent.
2469 Details : Ratelimit a memory allocation failure message that can
2474 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2475 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2479 Description: Liblustre build fixes.
2480 Details : Liblustre build fixes.
2484 Description: libcfs: add cfs_{need,cond}_resched() interface.
2485 Details : libcfs: add cfs_{need,cond}_resched() definition and
2486 implementations for Linux, NT, and liblustre.
2488 Severity : enhancement
2490 Description: More exported tunables for mballoc
2491 Details : Add support for tunable preallocation window and new tunables for
2492 large/small requests
2496 Description: Detect corruption of block bitmap and checking for preallocations
2497 Details : Checks validity of on-disk block bitmap. Also it does better
2498 checking of number of applied preallocations. When corruption is
2499 found, it turns filesystem readonly to prevent further corruptions.
2503 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2504 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2505 window to 0. If it does hit cache miss, and read-pattern is still
2506 stride-io mode, does not reset the stride window, but also does not
2507 increase the stride window length in this case.
2511 Frequency : only for big-endian servers
2512 Description: Check if system is big-endian while mounting fs with extents feature
2513 Details : Mounting a filesystem with extents feature will fail on big-endian
2514 systems since ext3-based ldiskfs is not supported on big-endian
2515 systems. This can be over-riden with "bigendian_extents" mount option.
2517 Severity : enhancement
2519 Description: The root squash functionality
2520 Details : A security feature, which is to prevent users from being able
2521 to mount lustre on their desktop, run as root, and delete
2522 all of the files in the filesystem. The goal is accomplished by
2523 remapping user id (UID) and group id (GID) of the root user to
2524 a UID and GID specified by the system administartor via Lustre
2525 configuration management server (MGS). The functionality also
2526 allows to specify sets of clients for which the remapping does
2531 Description: Excessive recovery window
2532 Details : With AT enabled, the recovery window can be excessively long (6000+
2533 seconds). To address this problem, we no longer use
2534 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2535 timeout no longer depends on the service time, it is set to
2536 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2537 time via pb_service_time.
2541 Description: Watchdog triggered on MDS failover
2542 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2543 the OSTs know that the MDS "UUID" can be reused for the same export
2544 from a different NID, so we do not need to wait for the export to be
2548 Frequency : rare, only if using MMP with Linux RAID
2550 Description: MMP doesn't work with Linux RAID
2551 Details : While using HA for Lustre servers with Linux RAID, it is possible
2552 that MMP will not detect multiple mounts. To make this work we
2553 need to unplug the device queue in RAID when the MMP block is being
2554 written. Also while reading the MMP block, we should read it from
2555 disk and not the cached one.
2557 Severity : enhancement
2559 Description: open file using fid
2560 Details : A file can be opened using just its fid, like
2561 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2564 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2567 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2568 Details : add proper locking for creation of jbd_4k slab cache
2572 Description: MMP check in ext3_remount() fails without displaying any error
2573 Details : When multiple mount protection fails during remount, proper error
2576 Severity : enhancement
2578 Description: Allow stripe size to be up to 4G-64k
2579 Details : Fix math logic to allow large stripe sizes.
2583 Description: add check for >8TB ldiskfs filesystems
2584 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2585 Don't allow >8TB ldiskfs filesystems to be mounted without
2586 force_over_8tb mount option
2591 Descriptoin: OSS complains "too many blocks" and crashes
2592 Details : In filter_commitrw_write():
2593 1. lquota_chkquota();
2594 2. obdo_from_inode();
2595 3. lquota_pending_commit();
2596 "2." has changed uid and gid in this case, then caused
2597 lquota_chkquota() and lquota_pending_commit() to use
2598 different uid and gid. That is the root of the bug.
2602 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2603 Details : Detect and print out the 'stripe count = -1' case properly.
2607 Description: Out or order replies might be lost on replay
2608 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2609 request with tid smaller then one currently being added, add it
2610 to the start, not end of the list.
2612 --------------------------------------------------------------------------------
2614 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2616 * Support for kernels:
2617 2.6.5-7.283 (SLES 9),
2618 2.6.9-55.EL (RHEL 4),
2619 2.6.16.46-0.14 (SLES 10),
2620 2.6.18.8 vanilla (kernel.org)
2621 * Client support for unpatched kernels:
2622 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2623 2.6.16 - 2.6.22 vanilla (kernel.org)
2624 * Due to recently discovered recovery problems, we do not recommend
2625 using patchless RHEL 4 clients with this or any earlier release.
2626 * Recommended e2fsprogs version: 1.39.cfs8
2627 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2628 * Starting with this release, the ldiskfs backing filesystem required
2629 by Lustre is now in its own package, lustre-ldiskfs. This package
2630 should be installed. It is versioned separately from Lustre and
2631 may be released separately in future.
2635 Description: block reactivating mgc import until all deactivates complete
2636 Details : Fix race when failing back MDT/MGS to itself (testing)
2639 Frequency : at statup only
2641 Description: mds_lov_synchronize race leads to various problems
2642 Details : simultaneous MDT->OST connections at startup can cause the
2643 sync to abort, leaving the OSC in a bad state.
2645 Severity : enhancement
2647 Description: add optional extra BUILD_VERSION info
2648 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2649 to override the lustre version.
2652 Frequency : 2.6.18 servers only
2654 Description: ll_kern_mount() doesn't release the module reference
2655 Details : The ldiskfs module reference count never drops down to 0
2656 because ll_kern_mount() doesn't release the module reference.
2661 Description: server LBUG when using old ost_num_threads parameter
2662 Details : Accept the old ost_num_threads parameter but warn that it
2663 is deprecated, and fix an off-by-one error that caused an LBUG.
2668 Description: Transient SCSI error results in persistent IO issue
2669 Details : iobuf->dr_error is not reinitialized to 0 between two
2673 Frequency : sometimes when underlying device returns I/O errors
2675 Description: OSTs not going read-only during write failures
2676 Details : OSTs are not remounted read-only when the journal commit threads
2677 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2678 instead of the ext3 wrappers.
2682 Description: poor connect scaling with increasing client count
2683 Details : Don't run filter_grant_sanity_check for more than 100 exports
2684 to improve scaling for large numbers of clients.
2687 Frequency : SLES10 only
2689 Description: sanity-quota.sh quotacheck failed: rc = -22
2690 Details : Quotas cannot be enabled on SLES10.
2693 Frequency : liblustre clients only
2695 Description: getdirentries does not give error when run on compute nodes
2696 Details : getdirentries does not fail when the size specified as an argument
2697 is too small to contain at least one entry
2699 Severity : enhancement
2701 Description: Add LNET router traceability for debug purposes
2702 Details : If a checksum failure occurs with a router as part of the
2703 IO path, the NID of the last router that forwarded the bulk data
2704 is printed so it can be identified.
2709 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2710 Details : Due to a race condition, liblustre clients were occasionally
2711 evicted incorrectly.
2713 Severity : enhancement
2715 Description: lfs setstripe use optional parameters instead of postional
2718 Severity : enhancement
2720 Description: Nanosecond timestamp support for ldiskfs
2721 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2722 resolution timestamps. There is not yet support for this at
2723 the Lustre filesystem level.
2726 Frequency : during server recovery
2728 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2729 Details : request with rq_no_resend flag not awake l_wait_event if they get
2733 Frequency : nfs export on patchless client
2735 Description: connectathon hang when test nfs export over patchless client
2736 Details : Disconnected dentry cannot be found with lookup, so we do not need
2737 to unhash it or make it invalid
2740 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2741 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2742 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2747 Description: open req refcounting wrong on reconnect
2748 Details : If reconnect happened between getting open reply from server and
2749 call to mdc_set_replay_data in ll_file_open, we will schedule
2750 replay for unreferenced request that we are about to free.
2751 Subsequent close will crash in variety of ways.
2752 Check that request is still eligible for replay in
2753 mdc_set_replay_data().
2758 Description: disable writes to filesystem when reading health_check file
2759 Details : the default for reading the health_check proc file has changed
2760 to NOT do a journal transaction and write to disk, because this
2761 can cause reads of the /proc file to hang and block HA state
2762 checking on a healthy but otherwise heavily loaded system. It
2763 is possible to return to the previous behaviour during configure
2764 with --enable-health-write.
2766 Severity : enhancement
2768 Description: 64-bit inode version
2769 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2770 to the inode. This will be required for version-based recovery.
2775 Description: MDS fails to start if a duplicate client export is detected
2776 Details : in some rare cases it was possible for a client to connect to
2777 an MDS multiple times. Upon recovery the MDS would detect this
2778 and fail during startup. Handle this more gracefully.
2780 Severity : enhancement
2782 Description: Add -o localflock option to simulate old noflock
2784 Details : This will achieve local-only flock/fcntl locks
2790 Description: log_commit_thread vs filter_destroy race leads to crash
2791 Details : Take import reference before releasing llog record semaphore
2796 Description: Wrong request locking in request set processing
2797 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2798 imp_delayed_list, in this place should be used imp_lock.
2801 Frequency : when reconnection
2803 Description: Grant Leak when osc reconnect to OST
2804 Details : When osc reconnect ost, OST(filter) should check whether it
2805 should grant more space to client by comparing fed_grant and
2806 cl_avail_grant, and return the granted space to client instead
2807 of "new granted" space, because client will call osc_init_grant
2808 to update the client grant space info.
2811 Frequency : when client reconnect to OST
2813 Description: Grant Leak when osc do resend and replay bulk write
2814 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2815 bulk write request, because the grant info while be sync between
2816 OSC and OST when reconnect, and we should ignore the grant info
2817 these of resend/replay write req.
2822 Description: Grant space more than avaiable left space sometimes.
2823 Details : When then OST is about to be full, if two bulk writing from
2824 different clients came to OST. Accord the avaliable space of the
2825 OST, the first req should be permitted, and the second one
2826 should be denied by ENOSPC. But if the seconde arrived before
2827 the first one is commited. The OST might wrongly permit second
2828 writing, which will cause grant space > avaiable space.
2831 Frequency : when client is evicted
2833 Description: Grant might be wrongly erased when osc is evicted by OST
2834 Details : when the import is evicted by server, it will fork another
2835 thread ptlrpc_invalidate_import_thread to invalidate the
2836 import, where the grant will be set to 0. While the original
2837 thread will update the grant it got when connecting. So if
2838 the former happened latter, the grant will be wrongly errased
2839 because of this race.
2844 Description: Checking Stale with correct fid
2845 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2846 to check whether it is stale, because sometimes, we want the
2847 enqueue happened anyway, and op_data.fid2 will not be initialized.
2849 Severity : enhancement
2851 Description: update patchless client
2852 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2855 Frequency : only with 2.4 kernel
2857 Description: random memory corruption
2858 Details : size of struct ll_inode_info is to big for union inode.u and this
2859 can be cause of random memory corruption.
2864 Description: Memory leak in recovery
2865 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2866 It should also check obd_fail before fsfilt_start, otherwise if
2867 fsfilt_start return -EROFS,(failover mds during mds recovery).
2868 then the req will return with repmsg->transno = 0 and rc = EROFS.
2869 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2870 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2871 be freed no matter whether fsfilt_commit success or not.
2874 Frequency : only with huge count clients
2876 Description: Prevents from taking the superblock lock in llap_from_page for
2878 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2879 for llap_from_page call in ll_removepage prevents from taking the
2880 superblock lock for a soon died page.
2885 Description: Not check open intent error before release open handle
2886 Details : in some rare cases, the open intent error is not checked before
2887 release open handle, which may cause
2888 ASSERTION(open_req->rq_transno != 0), because it tries to release
2889 the failed open handle.
2894 Description: Set cat log bitmap only after create log success.
2895 Details : in some rare cases, the cat log bitmap is set too early. and it
2896 should be set only after create log success.
2900 Description: Accessing a block bevice can re-enable I/O when Lustre is
2901 tearing down a device.
2902 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2907 Description: service threads may hog cpus when there are a lot of requests
2908 Details : Insert cond_resched to give other threads a chance to use some CPU
2913 Description: the cat log was not initialized in recovery
2914 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2915 unlink log on mds will not be initialized until mds post
2916 recovery. And also in mds post recovery, the unlink log will
2917 initialization will be done asynchronausly, so there will be race
2918 between add unlink log and unlink log initialization.
2922 Description: brw_stats were being printed incorrectly
2923 Details : brw_stats were being printed as log2 but all of them were not
2924 recorded as log2. Also remove some code duplication arising from
2925 filter_tally_{read,write}.
2929 Frequency : rare, only in recovery.
2930 Description: ASSERTION(req->rq_type != LI_POISON) failed
2931 Details : imp_lock should be held while iterating over imp_sending_list for
2932 prevent destroy request after get timeout in ptlrpc_queue_wait.
2936 Description: replay-single.sh test 52 fails
2937 Details : A lock's skiplist need to be cleanup when it being unlinked
2938 from its resource list.
2942 Description: Short directio read returns full requested size rather than
2944 Details : Direct I/O operations should return actual amount of bytes
2945 transferred rather than requested size.
2947 Severity : enhancement
2949 Description: metadata RPC reduction (e.g. for rm performance)
2950 Details : decrease the amount of synchronous RPC between clients and servers
2951 by canceling conflicing lock before the operation on the client
2952 and packing thier handles into the main operation RPC to server.
2954 Severity : enhancement
2956 Description: Async OSC create to avoid the blocking unnecessarily.
2957 Details : If a OST has no remain object, system will block on the creating
2958 when need to create a new object on this OST. Now, ways use
2959 pre-created objects when available, instead of blocking on an
2960 empty osc while others are not empty. If we must block, we block
2961 for the shortest possible period of time.
2965 Description: improve handling recoverable errors
2966 Details : if request processig with error which can be recoverable on server
2967 request should be resend, otherwise page released from cache and
2970 Severity : enhancement
2972 Description: refine locking for avoid write wrong info into lov_objid file
2973 Details : fix possible races with add new target and write/update data in
2976 --------------------------------------------------------------------------------
2978 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2983 Frequency : on some architectures
2985 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2986 Details : Uninitialized flags sometimes cause configuration commands to
2990 Frequency : patchless clients only
2992 Description: missing __iget() symbol export
2993 Details : The __iget() symbol export is missing. To avoid the need for
2994 this on patchless clients the deathrow inode reaper is turned
2995 off, and we depend on the VM to clean up old inodes. This
2996 dependency was during via the fix for bug 12181.
2998 --------------------------------------------------------------------------------
3000 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3002 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3003 INTEROPERATE with older versions automatically. In many cases a
3004 special upgrade step is needed. Please read the
3005 user documentation before upgrading any part of a live system.
3006 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3007 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3008 * WARNING: Lustre configuration and startup changes are required with
3009 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3013 Severity : enhancement
3015 Description: MountConf
3016 Details : Lustre configuration is now managed via mkfs and mount
3017 commands instead of lmc and lconf. New obd types (MGS, MGC)
3018 are added for dynamic configuration management. See
3019 https://mail.clusterfs.com/wikis/lustre/MountConf for
3022 Severity : enhancement
3024 Description: dynamic OST addition
3025 Details : OSTs can now be added to a live filesystem
3027 Severity : enhancement
3029 Description: startup order invariance
3030 Details : MDTs and OSTs can be started in any order. Clients only
3031 require the MDT to complete startup.
3033 Severity : enhancement
3035 Description: parallel, asynchronous orphan cleanup
3036 Details : orphan cleanup is now performed in separate threads for each
3037 OST, allowing parallel non-blocking operation.
3039 Severity : enhancement
3041 Description: optimized stripe assignment
3042 Details : stripe assignments are now made based on ost space available,
3043 ost previous usage, and OSS previous usage, in order to try
3044 to optimize storage space and networking resources.
3046 Severity : enhancement
3048 Description: Permanently set tunables
3049 Details : All writable /proc/fs/lustre tunables can now be permanently
3050 set on a per-server basis, at mkfs time or on a live system.
3052 Severity : enhancement
3054 Description: Lustre message v2
3055 Details : Add lustre message format v2.
3057 Severity : enhancement
3059 Description: client OST exclusion list
3060 Details : Clients can be started with a list of OSTs that should be
3061 declared "inactive" for known non-responsive OSTs.
3065 Description: SPEC SFS validation failure on NFS v2 over lustre.
3066 Details : Changes the blocksize for regular files to be 2x RPC size,
3067 and not depend on stripe size.
3069 Severity : enhancement
3071 Description: Multiple MD RPCs in flight.
3072 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3073 To avoid overly-overloading MDS, introduce a limit on number of
3074 MDS RPCs in flight for a single client and add /proc controls
3075 to adjust this limit.
3077 Severity : enhancement
3079 Description: client read/write statistics
3080 Details : Add client read/write call usage stats for performance
3081 analysis of user processes.
3082 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3083 file access. extents_stats shows chunk size distribution.
3084 extents_stats_per_process show chunk size distribution per
3087 Severity : enhancement
3089 Description: mds statistics
3090 Details : Add detailed mds operations statistics in
3091 /proc/fs/lustre/mds/*/stats.
3095 Description: Failure of copying files with lustre special EAs.
3096 Details : Client side always return success for setxattr call for lustre
3097 special xattr (currently only "trusted.lov").
3102 Description: Refcount LNET uuids
3103 Details : The global LNET uuid list grew linearly with every startup;
3104 refcount repeated list entries instead of always adding to
3107 Severity : enhancement
3109 Description: Dynamic service threads
3110 Details : Within a small range, start extra service threads
3111 automatically when the request queue builds up.
3114 Frequency : mixed-endian client/server environments
3116 Description: mixed-endian crashes
3117 Details : The new msg_v2 system had some failures in mixed-endian
3120 Severity : enhancement
3122 Description: Easy OST removal
3123 Details : OSTs can be permanently deactivated with e.g. 'lctl
3124 conf_param lustre-OST0001.osc.active=0'
3126 Severity : enhancement
3128 Description: MGS proc entries
3129 Details : Added basic proc entries for the MGS showing what filesystems
3132 Severity : enhancement
3134 Description: provide MGS failover
3135 Details : Added config lock reacquisition after MGS server failover.
3137 Severity : enhancement
3139 Description: add Linux 2.4 support
3140 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3144 Description: a large application tries to do I/O to the same resource and dies
3145 in the middle of it.
3146 Details : Check the req->rq_arrival time after the call to
3147 ost_brw_lock_get(), but before we do anything about
3148 processing it & sending the BULK transfer request. This
3149 should help move old stale pending locks off the queue as
3150 quickly as obd_timeout.
3153 Frequency : when an incorrect nid is specified during startup
3155 Description: ptlrpc connect to non-existant node causes kernel crash
3156 Details : LNET can't be re-entered from an event callback, which
3157 happened when we expire a message after the export has been
3158 cleaned up. Instead, hand the zombie cleanup off to another
3161 Severity : enhancement
3163 Description: plain/inodebits lock performance improvement
3164 Details : Grouping plain/inodebits in granted list by their request modes
3165 and bits policy, thus improving the performance of search through
3169 Frequency : only if OST filesystem is corrupted
3171 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3172 Details : for a short time RPCs with bulk IO are in the replay list,
3173 but replay of bulk IOs is unimplemented. If the OST filesystem
3174 is corrupted due to disk cache incoherency and then replay is
3175 started it is possible to trip an assertion. Avoid putting
3176 committed RPCs into the replay list at all to avoid this issue.
3179 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3181 Description: System hang on startup
3182 Details : This bug allowed the liblustre (e.g. catamount) client to
3183 return to the app before handling all startup RPCs. This
3184 could leave the node unresponsive to lustre network traffic
3185 and manifested as a server ptllnd timeout.
3187 Severity : enhancement
3189 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3190 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3191 boolean to control whether to print peer debug info when a
3192 client's RPC times out.
3195 Frequency : only for kernels with patches from Lustre below 1.4.3
3197 Description: Remove old rdonly API
3198 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3201 Frequency : only for devices with external journals
3203 Description: Set external device read-only also
3204 Details : During a commanded failover stop, we set the disk device
3205 read-only while the server shuts down. We now also set any
3206 external journal device read-only at the same time.
3209 Frequency : when upgrading from 1.4 while trying to change parameters
3211 Description: The wrong (new) MDC name was used when setting parameters for
3212 upgraded MDT's. Also allows changing of OSC (and MDC)
3213 parameters if --writeconf is specified at tunefs upgrade time.
3216 Frequency : when setting specific ost indicies
3218 Description: QOS code breaks on skipped indicies
3219 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3220 created with --index need not be sequential.
3222 Severity : enhancement
3224 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3225 Details : The uninit_groups feature works in conjunction with the kernel
3226 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3227 up the pass1 processing of e2fsck. This is a read-only feature
3228 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3229 mount filesystems that have had this feature enabled.
3231 Severity : enhancement
3233 Description: Improve multi-block allocation algorithm to avoid fragmentation
3234 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3235 allocation locality and avoid filesystem fragmentation.
3237 ------------------------------------------------------------------------------
3239 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3241 * Support for kernels:
3242 2.4.21-47.0.1.EL (RHEL 3)
3243 2.6.5-7.283 (SLES 9)
3244 2.6.9-42.0.10.EL (RHEL 4)
3245 2.6.12.6 vanilla (kernel.org)
3246 2.6.16.27-0.9 (SLES 10)
3247 * Recommended e2fsprogs version: 1.39.cfs5
3249 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3253 Frequency : occasional, depends on client load and configuration
3254 Bugzilla : 12181, 12203
3255 Description: data loss for recently-modified files
3257 Details : In some cases it is possible that recently written or created
3258 files may not be written to disk in a timely manner (this should
3259 normally be within 30s unless client IO load is very high).
3260 The problem appears as zero-length files or files that are a
3261 multiple of 1MB in size after a client crash or client eviction
3262 that are missing data at the end of the file.
3264 This problem is more likely to be hit on clients where files are
3265 repeatedly created and unlinked in the same directory, clients
3266 have a large amount of RAM, have many CPUs, the filesystem has
3267 many OSTs, the clients are rebooted frequently, and/or the files
3268 are not accessed by other nodes after being written.
3270 The presence of the problem can be detected by looking at
3271 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3272 smaller than the second (nr_unused) then dirty files will not
3273 be flushed automatically to disk. "sync; sleep 10" should be
3274 run several times on the node before unmounting it to update
3275 Lustre (this is also safe to run on nodes without this problem).
3277 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3278 that can cause this same problem, so customers using that kernel
3279 also need to update the kernel in addition to Lustre. In order
3280 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3282 It is normal that files written just before a client crash (less
3283 than 30s) may not yet have been flushed to disk, even for local
3287 Frequency : frequent on thin XT3 nodes
3289 Description: UUID collision on thin XT3 Linux nodes
3290 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3291 randomly, since we relied on an insufficiently-seeded PRNG.
3296 Description: OSS hangs after "All ost request buffers busy"
3297 Details : A deadlock between quota and journal operations caused OSS
3298 hangs after printing "All ost request buffers busy."
3301 Frequency : always on liblustre builds
3303 Description: Cleanup compiler warnings on liblustre
3306 Frequency : always on liblustre builds on XT3
3308 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3312 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3313 > 32000 subdirectories
3314 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3315 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3316 subdirectory count crosses 32000. This will aid e2fsck to
3317 correctly handle more than 32000 subdirectories.
3320 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3322 Description: System hang on startup
3323 Details : This bug allowed the liblustre (e.g. catamount) client to
3324 return to the app before handling all startup RPCs. This
3325 could leave the node unresponsive to lustre network traffic
3326 and manifested as a server ptllnd timeout.
3328 Severity : enhancement
3330 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3331 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3332 boolean to control whether to print peer debug info when a
3333 client's RPC times out.
3338 Description: make O_SYNC working on 2.6 kernels
3339 Details : 2.6 kernels use different method for mark pages for write,
3340 so need add a code to lustre for O_SYNC work.
3345 Description: Failure to close file and release space on NFS
3346 Details : Put inode details into lock acquired in ll_intent_file_open.
3347 Use mdc_intent_lock in ll_intent_open to properly
3348 detect all kind of errors unhandled by mdc_enqueue.
3353 Description: proc file read during shutdown sometimes raced obd removal,
3355 Details : Add lock to prevent obd access after proc file removal.
3358 Frequency : Only for files larger than 4GB on 32-bit clients.
3360 Description: improperly doing page alignment of locks
3361 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3362 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3367 Description: RPCs being resent when they shouldn't be
3368 Details : Some RPCs that should not be resent are being resent. This
3369 can cause inconsistencies in the RPC state machine. Do not
3370 resend such requests.
3373 Frequency : rare, only with NFS export
3375 Description: Crash on NFS re-export node
3376 Details : under very unusual load conditions an assertion is hit in
3377 ll_intent_file_open()
3380 Frequency : only if OST filesystem is corrupted
3382 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3383 Details : for a short time RPCs with bulk IO are in the replay list,
3384 but replay of bulk IOs is unimplemented. If the OST filesystem
3385 is corrupted due to disk cache incoherency and then replay is
3386 started it is possible to trip an assertion. Avoid putting
3387 committed RPCs into the replay list at all to avoid this issue.
3392 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3393 Details : Large single O_DIRECT read and write calls can fail to allocate
3394 a sufficiently large buffer to process the request. In case of
3395 allocation failure the allocation is retried with a smaller
3396 buffer and broken into smaller requests.
3398 Severity : enhancement
3400 Description: Add -o localflock option to simulate old noflock behaviour.
3401 Details : This will achieve local-only flock/fcntl locks coherentness.
3406 Description: versioning check is incomplete
3407 Details : Checking the version difference of client vs. server, report
3408 error if the gap is too big.
3413 Description: add support PG_writeback bit
3414 Details : add support for PG_writeback bit for Lustre, for more carefull
3415 work with page cache in 2.6 kernel. This also fix some deadlocks
3416 and remove hack for work O_SYNC with 2.6 kernel.
3418 Severity : enhancement
3420 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3421 Details : The uninit_groups feature works in conjunction with the kernel
3422 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3423 up the pass1 processing of e2fsck. This is a read-only feature
3424 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3425 mount filesystems that have had this feature enabled.
3427 Severity : enhancement
3429 Description: Improve multi-block allocation algorithm to avoid fragmentation
3430 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3431 allocation locality and avoid filesystem fragmentation.
3433 ------------------------------------------------------------------------------
3435 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3437 * Support for kernels:
3438 2.6.9-42.0.3.EL (RHEL 4)
3439 2.6.5-7.276 (SLES 9)
3440 2.4.21-47.0.1.EL (RHEL 3)
3441 2.6.12.6 vanilla (kernel.org)
3442 2.6.16.21-0.8 (SLES10)
3443 * Recommended e2fsprogs version: 1.39.cfs2-0
3445 * The backwards-compatible /proc/sys/portals symlink has been removed
3446 in this release. Before upgrading, please ensure that you change
3447 any configuration scripts or /etc/sysctl.conf files that access
3448 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3449 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3450 in advance of the upgrade on any system running Lustre 1.4.6 or
3451 newer, since /proc/sys/lnet was added in that version.
3452 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3456 Frequency : only when quota is used
3458 Description: avoid scanning export list for quota master
3459 Details : Change the algorithms to avoid scanning export list in order
3460 to improve the efficiency.
3463 Frequency : MDS failover only, very rarely
3465 Description: "went back in time" messages on mds failover
3466 Details : The greatest transno may be lost when the current operation
3467 finishes with an error (transno==0) and the client's last_rcvd
3468 record is over-written. Save the greatest transno in the
3469 mds_last_transno for this case.
3472 Frequency : always for specific kernels and striping counts
3474 Description: client may get "Matching packet too big" without ACL support
3475 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3476 when trying to access files in certain configurations. The
3477 clients should in fact be denied when mounting because they do
3478 not understand ACLs.
3481 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3483 Description: many clients connecting with IO in progress causes connect timeouts
3484 Details : Avoid synchronous journal commits to avoid delays caused by many
3485 clients connecting/disconnecting when bulk IO is in progress.
3486 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3487 OST_IO_PORTAL to avoid delays behind potentially many pending
3491 Frequency : occasionally with multiple writers to a single file
3493 Description: shared writes to file may result in wrong size reported by stat()
3494 Details : Allow growing of kms when extent lock is cancelled
3497 Frequency : always with random mmap IO to multi-striped file
3499 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3500 Details : Only if the hole is at the end of OST object so that kms is too
3501 small. Fix is to increase kms accordingly in ll_nopage.
3504 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3506 Description: writes to a missing object would leak memory on the OST
3507 Details : If there is an inconsistency between the MDS and OST filesystems,
3508 such that the MDS references an object that doesn't exist, writes
3509 to that object will leak memory due to incorrect cleanup in the
3510 error handling path, eventually running out of memory on the OST.
3515 Description: Creating too long symlink causes lustre errors
3516 Details : Check symlink and name lengths before sending requests to MDS.
3519 Frequency : only if flock is enabled (not on by default)
3521 Description: posix locks not released on fd closure on 2.6.9+
3522 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3523 kernels, this caused such locks not to be released on fd close and
3524 then assertions on fs unmount about still used locks.
3527 Frequency : MDS failover only, very rarely
3529 Description: clients may get ASSERTION(granted_lock != NULL)
3530 Details : When request was taking a long time, and a client was resending
3531 a getattr by name lock request. The were multiple lock requests
3532 with the same client lock handle and
3533 mds_getattr_name->fixup_handle_for_resent_request found one of the
3534 lock handles but later failed with ASSERTION(granted_lock != NULL).
3539 Description: handle->h_buffer_credits > 0, assertion failure
3540 Details : h_buffer_credits is zero after truncate, causing assertion
3541 failure. This patch extends the transaction or creates a new
3545 Frequency : NFS re-export or patchless client
3546 Bugzilla : 11179, 10796
3547 Description: Crash on NFS re-export node (__d_move)
3548 Details : We do not want to hash the dentry if we don't have a lock.
3549 But if this dentry is later used in d_move, we'd hit uninitialised
3550 list head d_hash, so we just do this to init d_hash field but
3551 leave dentry unhashed.
3554 Frequency : NFS re-export or patchless client
3556 Description: NFS exports has problem with symbolic link
3557 Details : lustre client didn't properly install dentry when re-exported
3558 to NFS or running patchless client.
3561 Frequency : NFS re-export or patchless client
3563 Description: Various nfs/patchless fixes.
3564 Details : fixes reuse disconected alias for lookup process - this fixes
3565 warning "find_exported_dentry: npd != pd",
3566 fix permission error with open files at nfs.
3567 fix apply umask when do revalidate.
3570 Frequency : occasional
3572 Description: Crash on NFS re-export node
3573 Details : calling clear_page() on the wrong pointer triggered oops in
3574 generic_mapping_read().
3577 Frequency : rarely, using O_DIRECT IO
3579 Description: unaligned directio crashes client with LASSERT
3580 Details : check for unaligned buffers before trying any requests.
3583 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3585 Description: stale data returned from RAID cache
3586 Details : If only a small amount of IO is done to the RAID device before
3587 reading it again it is possible to get stale data from the RAID
3588 cache instead of reading it from disk.
3591 Frequency : always for sles10 kernel
3593 Description: sles10 support
3594 Details : ll_follow_link: compile fixes and using of nd_set_link
3595 under newer kernels.
3598 Frequency : depends on arch, kernel and compiler version, always on sles10
3601 Description: recursive or deep enough symlinks cause stack overflow
3602 Details : getting rid of large stack-allocated variable in
3606 Frequency : depends on hardware
3608 Description: lustre write performance loss in the SLES10 kernel
3609 Details : the performance loss is caused by using of write barriers in the
3610 ext3 code. The SLES10 kernel turns barrier support on by
3611 default. The fix is to undo that change for ldiskfs.
3613 ------------------------------------------------------------------------------
3615 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3617 * Support for kernels:
3618 2.6.9-42.0.3EL (RHEL 4)
3619 2.6.5-7.276 (SLES 9)
3620 2.4.21-47.0.1.EL (RHEL 3)
3621 2.6.12.6 vanilla (kernel.org)
3625 Frequency : quota enabled and large files being deleted
3627 Description: releasing more than 4GB of quota at once hangs OST
3628 Details : If a user deletes more than 4GB of files on a single OST it
3629 will cause the OST to spin in an infinite loop. Release
3630 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3635 Description: statfs data retrieved from /proc may be stale or zero
3636 Details : When reading per-device statfs data from /proc, in the
3637 {kbytes,files}_{total,free,avail} files, it may appear
3638 as zero or be out of date.
3641 Frequency : systems with MD RAID1 external journal devices
3643 Description: lconf's call to blkid is confused by RAID1 journal devices
3644 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3645 of returning all block devices that match the journal UUID.
3648 Frequency : always, for aggregate stripe size over 4GB
3650 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3651 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3652 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3655 Frequency : always if "lfs find" used on a local file/directory
3657 Description: "lfs find" segfaults if used on a local file/directory
3658 Details : The case where a directory component was not specified wasn't
3659 handled correctly. Handle this properly.
3662 Frequency : always on ppc64
3664 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3665 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3666 perform endianness conversion.
3669 Frequency : rarely (truncate to non-zero file size after write under load)
3670 Bugzilla : 10730, 10687
3671 Description: Files padded with zeros to next 4K multiple
3672 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3673 it is possible that files that are truncated to a non-zero size
3674 immediately after being written are filled with zero bytes beyond
3675 the truncated size. No file data is lost.
3677 Severity : enhancement
3679 Description: Allow recovery/failover for liblustre clients.
3680 Details : liblustre clients were unaware of failover configurations until
3683 Severity : enhancement
3685 Description: user file locks should fail when not mounting with flock option
3686 Details : Set up an error-returning stub in ll_file_operations.lock field
3687 to prevent incorrect behaviour when client is mounted without
3688 flock option. Also, set up properly f_op->flock field for
3692 Frequency : always on ia64
3694 Description: "lfs df" loops on printing out MDS statfs information
3695 Details : The obd_ioctl_data was not initialized and in some systems
3696 this caused a failure during the ioctl that did not return
3697 an error. Initialize the struct and return an error on failure.
3700 Frequency : SLES 9 only
3702 Description: Error of copying files with lustre special EAs as root
3703 Details : Client side always return success for setxattr call for lustre
3704 special xattr (currently only "trusted.lov").
3707 Frequency : rarely on clusters with both ia64+i386 clients
3709 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3710 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3711 clients are doing concurrent writes to the same file it is
3712 possible that the ia64 clients may LASSERT because the OST
3713 extent locks are not PAGE_SIZE aligned. Ensure that grown
3714 locks are always aligned on the request boundary.
3717 Frequency : specific use, occasional
3719 Description: Overwriting in use executable truncates on-disk binary image
3720 Details : If one node attempts to overwrite an executable in use by
3721 another node, we now correctly return ETXTBSY instead of
3722 truncating the file.
3724 Severity : enhancement
3726 Description: Async OSC create to avoid the blocking unnecessarily.
3727 Details : If a OST has no remain object, system will block on the creating
3728 when need to create a new object on this OST. Now, ways use
3729 pre-created objects when available, instead of blocking on an
3730 empty osc while others are not empty. If we must block, we block
3731 for the shortest possible period of time.
3736 Description: chmod on Lustre root is propagated to other clients
3737 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3738 invalid by the follow_mount time.
3741 Frequency : liblustre clients only
3743 Description: Race in 'instant cancel' lock handling could lead to such locks
3744 never to be granted in case of SMP MDS
3745 Details : Do not destroy not yet granted but cbpending locks in
3749 Frequency : replay/resend of open
3751 Description: non null lock assetion failure in mds_intent_policy
3752 Details : Trying to replay/resend lockless open requests resulted in
3753 mds_open() returning 0 with no lock. Now it sets a flag if
3754 a lock is going to be returned.
3756 Severity : enhancement
3758 Description: Checksum enhancements
3759 Details : New checksum enhancements allow for resending RPCs that failed
3762 Severity : enhancement
3764 Description: Tunables on number of dirty pages in cacche
3765 Details : Allow to set limit on number of dirty pages cached.
3770 Description: client crash on unmount - lock still has references
3771 Details : In some error handling cases it was possible to leak a lock
3772 reference on a client while accessing a file. This was not
3773 harmful to the client during operation, but would cause the
3774 client to crash when the filesystem is unmounted.
3777 Frequency : specific case, rare
3779 Description: ETXTBSY on mds though file not in use by client
3780 Details : ETXTBSY is no longer incorrectly returned when attempting to
3781 chmod or chown a directory that the user previously tried to
3782 execute or a currently-executing binary.
3785 Frequency : extremely rare except on liblustre-based clients
3787 Description: Lustre space not freed when files are deleted
3788 Details : Clean up open-unlinked files after client eviction. Previously
3789 the unlink was skipped and the files remained as orphans.
3794 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3795 Details : In some cases it was possible to send a blocking callback to
3796 a client doing a glimpse, even though that client didn't get
3797 a lock granted. When the glimpse lock is cancelled on the OST
3798 the freed lock is left on the waiting list and corrupted the list.
3801 Frequency : all core dumps
3803 Description: Broke core dumps to lustre
3804 Details : Negative dentry may be unhashed if parent does not have UPDATE
3805 lock, but some callers, e.g. do_coredump, expect dentry to be
3806 hashed after successful create, hash it in ll_create_it.
3808 ------------------------------------------------------------------------------
3810 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3812 * Support for kernels:
3813 2.6.9-42.0.2.EL (RHEL 4)
3814 2.6.5-7.276 (SLES 9)
3815 2.4.21-40.EL (RHEL 3)
3816 2.6.12.6 vanilla (kernel.org)
3820 Frequency : always on RHEL 3
3822 Description: Number of open files grows over time
3823 Details : The number of open files grows over time, whether or not
3824 Lustre is started. This was due to a filp leak introduced
3825 by one of our kernel patches.
3827 ------------------------------------------------------------------------------
3829 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3831 * Support for kernels:
3832 2.6.9-42.EL (RHEL 4)
3833 2.6.5-7.276 (SLES 9)
3834 2.4.21-40.EL (RHEL 3)
3835 2.6.12.6 vanilla (kernel.org)
3840 Bugzilla : 5719, 9635, 9792, 9684
3841 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3842 Details : If a server is under heavy load and cannot reply to new
3843 connection requests before the client resends the (re)connect,
3844 the connection handling code can behave badly if two service
3845 threads are concurrently handing separate (re)connections from
3846 the same client. Add better locking to the connection handling
3847 code, and ensure that only a single connection will be processed
3848 for a given client UUID, even if the lock is dropped.
3850 Severity : enhancement
3852 Description: add TCP zero-copy support to kernel
3853 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3854 sends if the hardware supports scatter-gather and checksumming.
3855 This allows socklnd to do client-write and server-read more
3856 efficiently and reduce CPU utilization from skbuf copying.
3859 Frequency : only if NFS exporting from client
3861 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3862 Details : If setting attributes on a file created under NFS that had
3863 never been opened it would be possible to oops the client
3864 if the file had no objects.
3867 Frequency : always for liblustre
3869 Description: liblustre client does MDS+OSTs setattr RPC for each write
3870 Details : When doing a write from a liblustre client, the client
3871 incorrectly issued an RPC to the MDS and each OST the file was
3872 striped over in order to update the timestamps. When writing
3873 with small chunks and many clients this could overwhelm the MDS
3874 with RPCs. In all cases it would slow down the write because
3875 these RPCs are unnecessary.
3877 Severity : enhancement
3879 Description: allow number of MDS service threads to be changed at module load
3880 Details : It is now possible to change the number of MDS service threads
3881 running. Adding "options mds mds_num_threads={N}" to the MDS's
3882 /etc/modprobe.conf will set the number of threads for the next
3883 time Lustre is restarted (assuming the "mds" module is also
3884 reloaded at that time). The default number of threads will
3885 stay the same, 32 for most systems.
3890 Description: OST crash if filesystem is unformatted or corrupt
3891 Details : If an OST is started on a device that has never been formatted
3892 or if the filesystem is corrupt and cannot even mount then the
3893 error handling cleanup routines would dereference a NULL pointer.
3898 Description: NULL pointer deref in llap_from_page.
3899 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3900 incorrect) mapping if the page was truncated/reclaimed while it was
3901 searched for. Check for this condition and skip such pages when
3902 doing readahead. Introduce extra check to llap_from_page() to
3903 verify page->mapping->host is non-NULL (so page is not anonymous).
3906 Frequency : Sometimes when using sys_sendfile
3908 Description: "page not covered by a lock" warnings from ll_readpage
3909 Details : sendfile called ll_readpage without right page locks present.
3910 Now we introduced ll_file_sendfile that does necessary locking
3911 around call to generic_file_sendfile() much like we do in
3915 Frequency : with certain MDS communication failures at client mount time
3917 Description: NULL pointer deref after failed client mount
3918 Details : a client connection request may delayed by the network layer
3919 and not be sent until after the PTLRPC layer has timed out the
3920 request. If the client fails the mount immediately it will try
3921 to clean up before the network times out the request. Add a
3922 reference from the request import to the obd device and delay
3923 the cleanup until the network drops the request.
3926 Frequency : occasionally during client (re)connect
3928 Description: assertion failure during client (re)connect
3929 Details : processing a client connection request may be delayed by the
3930 client or server longer than the client connect timeout. This
3931 causes the client to resend the connection request. If the
3932 original connection request is replied in this interval, the
3933 client may trip an assertion failure in ptlrpc_connect_interpret()
3934 which thought it would be the only running connect process.
3937 Frequency : only with obd_echo servers and clients that are rebooted
3939 Description: kernel BUG accessing uninitialized data structure
3940 Details : When running an obd_echo server it did not start the ping_evictor
3941 thread, and when a client was evicted an uninitialized data
3942 structure was accessed. Start the ping_evictor in the RPC
3943 service startup instead of the OBD startup.
3945 Severity : enhancement
3946 Bugzilla : 10193 (patchless)
3947 Description: Remove dependency on various unexported kernel interfaces.
3948 Details : No longer need reparent_to_init, exit_mm, exit_files,
3949 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3952 Frequency : rare (only users of deprecated and unsupported LDAP config)
3954 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3955 Details : LDAP apparently contains 'lustreName' attributes instead of
3956 'name'. A simple remapping of the name is sufficient.
3959 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3961 Description: waiting_locks_callback trips kernel BUG if client is evicted
3962 Details : Running with the dump_on_timeout debug flag turned on makes
3963 it possible that the waiting_locks_callback() can try to dump
3964 the Lustre kernel debug logs from an interrupt handler. Defer
3965 this log dumping to the expired_lock_main() thread.
3967 Severity : enhancement
3969 Description: Support NFS exporting on 2.6 kernels.
3970 Details : Implement non-rawops metadata methods for NFS server to use without
3971 changing NFS server code.
3974 Frequency : very rare (synthetic metadata workload only)
3976 Description: two racing renames might cause an MDS thread to deadlock
3977 Details : Running the "racer" program may cause one MDS thread to rename
3978 a file from being the source of a rename to being the target of
3979 a rename at exactly the same time that another thread is doing
3980 so, and the second thread has already enqueued these locks after
3981 doing a lookup of the target and is trying to relock them in
3982 order. Ensure that we don't try to re-lock the same resource.
3985 Frequency : only very large systems with liblustre clients
3987 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3988 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3989 all OSTs in parallel. This allows the eviction of stale liblustre
3990 clients to proceed much faster than if they were done in series,
3991 and also offers similar improvements for other set_info RPCs.
3996 Description: excessive CPU usage during initial read phase on client
3997 Details : During the initial read phase on a client, it would agressively
3998 retry readahead on the file, consuming too much CPU and impacting
3999 performance (since 1.4.5.8). Improve the readahead algorithm
4000 to avoid this, and also improve some other common cases (read
4001 of small files in particular, where "small" is files smaller than
4002 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4007 Description: MDS crash when receiving packet with unknown intent.
4008 Details : Do not LBUG in unknown intent case, just return -EFAULT
4010 Severity : enhancement
4011 Bugzilla : 9293, 9385
4012 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4013 Details : Do not serialize getattr (non-intent version) and statfs.
4016 Frequency : occasional, when OST network is overloaded/intermittent
4018 Description: client evicted by OST after bulk IO timeout
4019 Details : If a client sends a bulk IO request (read or write) the OST
4020 may evict the client if it is unresposive to its data GET/PUT
4021 request. This is incorrect if the network is overloaded (takes
4022 too long to transfer the RPC data) or dropped the OST GET/PUT
4023 request. There is no need to evict the client at all, since
4024 the pinger and/or lock callbacks will handle this, and the
4025 client can restart the bulk request.
4028 Frequency : Always when mmapping file with no objects
4030 Description: client crashes when mmapping file with no objects
4031 Details : Check that we actually have objects in a file before doing any
4032 operations on objects in ll_vm_open, ll_vm_close and
4038 Description: Request leak when working with deleted CWD
4039 Details : Introduce advanced request refcount tracking for requests
4040 referenced from lustre intent.
4042 Severity : Enhancement
4044 Description: Cache open file handles on client.
4045 Details : MDS now will return special lock along with openhandle, if
4046 requested and client is allowed to hold openhandle, even if unused,
4047 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4048 closing files for every read/write openration.
4050 Severity : Enhancement
4052 Description: Cache open negative dentries on client when possible.
4053 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4054 negative dentries on lock revocation.
4059 Description: Remounting a client read-only wasn't possible with a zconf mount
4060 Details : It wasn't possible to remount a client read-only with llmount.
4062 Severity : enhancement
4063 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4064 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4065 MPICH is included to add Lustre-specific ADIO interfaces.
4066 This is based closely on the UFS ADIO layer and only differs
4067 in file creation, in order to allow the OST striping to be set.
4068 This is user-contributed code and not supported by CFS.
4073 Description: extended inode attributes (immutable, append-only) work improperly
4074 when 2.4 and 2.6 kernels are used on client/server or vice versa
4075 Details : Introduce kernel-independent values for these flags.
4077 Severity : enhancement
4080 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4081 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4082 granularity. Now, Fractional MB granularity is be supported,
4083 this is very useful for low memory system.
4085 Severity : enhancement
4087 Description: Getattr by fid
4088 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4089 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4090 have locks and avoids getattr rpc requests for it on every stat.
4093 Frequency : Always, for filesystems larger than 2TB
4095 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4096 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4097 filesystems larger than 2TB to be mounted with mballoc enabled.
4100 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4102 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4103 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4104 filesystems larger than 2TB is mounted then the kernel will
4105 silently corrupt the start of the filesystem. CONFIG_LBD is
4106 enabled for all CFS-supported kernels, but the possibility of
4107 this happening with a modified kernel config exists.
4109 Severity : enhancement
4111 Description: add client O_DIRECT support for 2.6 kernels
4112 Details : It is now possible to do O_DIRECT reads and writes to files
4113 in the Lustre client mountpoint on 2.6 kernel clients.
4115 Severity : enhancement
4117 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4118 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4119 parallel, not waiting for response from every OST before sending
4120 a rpc to the next OST.
4125 Description: setattr vs write race when updating file timestamps
4126 Details : Client processes that update a file timestamp into the past
4127 right after writing to the file (e.g. tar) it is possible that
4128 the updated file modification time can be reset to the current
4129 time due to a race between processing the setattr and write RPC.
4131 Severity : enhancement
4133 Description: Bring 'lfs find' closer in line with regular Linux find.
4134 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4135 -print0 options and obtains all the needed info through the lustre
4138 Severity : enhancement
4140 Description: support up to 1024 configured devices on one node
4141 Details : change obd_dev array from statically allocated to dynamically
4142 allocated structs as they are first used to reduce memory usage
4147 Description: Flush dirty partially truncated pages during truncate
4148 Details : Immediatelly flush partially truncated pages in filter_setattr,
4149 this way we completely avoid having any pages in page cache on OST
4150 and can retire ugly workarounds during writes to flush such pages.
4155 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4156 Details : protect inode from truncation within vfs_unlink() context
4157 just take a reference before calling vfs_unlink() and release it
4158 when parent's i_sem is free.
4163 Description: last_id value checked outside lock on OST caused LASSERT failure
4164 Details : If there were multiple MDS->OST object precreate requests in
4165 flight, it was possible that the OST's last object id was checked
4166 outside a lock and incorrectly tripped an assertion. Move checks
4167 inside locks, and discard old precreate requests.
4170 Frequency : always, if extents are used on OSTs
4172 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4173 Details : OSTs using the extents format would not zero the high 16 bits of
4174 the index physical block number. This is not a problem for any
4175 OST filesystems smaller than 16TB, and no kernels support ext3
4176 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4177 new/modified files) and can be fixed for existing filesystems
4178 with e2fsprogs-1.39-cfs1.
4183 Description: import connection selection may be incorrect if timer wraps
4184 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4185 import connections to be ignored if the 32-bit jiffies counter
4186 wraps. Use a 64-bit jiffies counter.
4189 Frequency : very large clusters immediately after boot
4191 Description: LNET request buffers exhausted under heavy short-term load
4192 Details : If a large number of client requests are generated on a service
4193 that has previously never seen so many requests it is possible
4194 that the request buffer growth cannot keep up with the spike in
4195 demand. Instead of dropping incoming requests, they are held in
4196 the LND until the RPC service can accept more requests.
4199 Frequency : Sometimes during replay
4201 Description: Assertion failure in ll_local_open after replay.
4202 Details : If replay happened on an open request reply before we were able
4203 to set replay handler, reply will become not swabbed tripping the
4204 assertion in ll_local_open. Now we set the handler right after
4205 recognising of open request
4208 Frequency : very rare
4210 Description: kernel reports "badness in vsnprintf"
4211 Details : Reading from the "recovery_status" /proc file in small chunks
4212 may cause a negative length in lprocfs_obd_rd_recovery_status()
4213 call to vsnprintf() (which is otherwise harmless). Exit early
4214 if there is no more space in the output buffer.
4216 Severity : enhancement
4218 Description: clear OBD RPC statistics by writing to them
4219 Details : It is now possible to clear the OBD RPC statistics by writing
4220 to the "stats" file.
4225 Description: Client mtime is not the same on different clients after utimes
4226 Details : In some cases, the client was using the utimes() syscall on
4227 a file cached on another node. The clients now validate the
4228 ctime from the MDS + OSTs to determine which one is right.
4233 Description: Inability to activate failout mode
4234 Details : lconf script incorrectly assumed that in pythong string's numeric
4235 value is used in comparisons.
4238 Frequency : always with multiple stripes per file
4240 Description: Inefficient object allocation for mutli-stripe files
4241 Details : When selecting which OSTs to stripe files over, for files with
4242 a stripe count that divides evenly into the number of OSTs,
4243 the MDS is always picking the same starting OST for each file.
4244 Return the OST selection heuristic to the original design.
4249 Description: mount failures may take full timeout to return an error
4250 Details : Under some heavy load conditions it is possible that a
4251 failed mount can wait for the full obd_timeout interval,
4252 possibly several minutes, before reporting an error.
4253 Instead return an error as soon as the status is known.
4255 Frequency : quota enabled and large files being deleted
4257 Description: releasing more than 4GB of quota at once hangs OST
4258 Details : If a user deletes more than 4GB of files on a single OST it
4259 will cause the OST to spin in an infinite loop. Release
4260 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4265 Description: statfs data retrieved from /proc may be stale or zero
4266 Details : When reading per-device statfs data from /proc, in the
4267 {kbytes,files}_{total,free,avail} files, it may appear
4268 as zero or be out of date.
4271 Frequency : systems with MD RAID1 external journal devices
4273 Description: lconf's call to blkid is confused by RAID1 journal devices
4274 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4275 of returning all block devices that match the journal UUID.
4278 Frequency : always, for aggregate stripe size over 4GB
4280 Description: assertion fails when trying to use 4GB stripe size
4281 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4282 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4285 Frequency : always on ppc64
4287 Description: the first write on an ext3 filesystem with mballoc got stuck
4288 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4289 perform endianness conversion.
4291 ------------------------------------------------------------------------------
4293 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4295 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4296 INTEROPERATE with older versions automatically. Please read the
4297 user documentation before upgrading any part of a live system.
4298 * WARNING: Lustre networking configuration changes are required with
4299 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4302 * Support for kernels:
4303 2.6.9-22.0.2.EL (RHEL 4)
4304 2.6.5-7.244 (SLES 9)
4305 2.6.12.6 vanilla (kernel.org)
4308 Severity : enhancement
4309 Bugzilla : 7981/8208
4310 Description: Introduced Lustre Networking (LNET)
4311 Details : LNET is new networking infrastructure for Lustre, it includes
4312 a reorganized network configuration mode (see the user
4313 documentation for full details) as well as support for routing
4314 between different network fabrics. Lustre Networking Devices
4315 (LNDs) for the supported network fabrics have also been
4316 created for this new infrastructure.
4318 Severity : enhancement
4319 Description: Introduced Access control lists
4320 Details : clients can set ACLs on files and directories in order to have
4321 more fine-grained permissions than the standard Unix UGO+RWX.
4322 The MDS must be started with the "-o acl" mount option.
4324 Severity : enhancement
4325 Description: Introduced filesystem quotas
4326 Details : Administrators may now establish per-user quotas on the
4329 Severity : enhancement
4331 Description: Configuration change for the XT3
4332 The PTLLND is now used to run Lustre over Portals on the XT3
4333 The configure option(s) --with-cray-portals are no longer used.
4334 Rather --with-portals=<path-to-portals-includes> is used to
4335 enable building on the XT3. In addition to enable XT3 specific
4336 features the option --enable-cray-xt3 must be used.
4341 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4342 Details : A race between allocating a new llcd and re-getting the llcd_lock
4343 allowed another thread to grab newly-allocated llcd.
4345 Severity : enhancement
4347 Description: 2.6 OST async journal commit and locking fix to improve performance
4348 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4349 2.6 kernels are now async as they already were in 2.4 kernels so
4350 that they can commit concurrently with the network bulk transfer.
4351 For block-allocated files the filter allocation semaphore is held
4352 to avoid filesystem fragmentation during allocation. BKL lock
4353 removed for 2.6 xattr operations where it is no longer needed.
4358 Description: lconf incorrectly determined whether two IP networks could talk
4359 Details : In some more complicated routing and multiple-network
4360 configurations, lconf will avoid trying to make a network
4361 connection to a disjoint part of the IP space. It was doing the
4362 math incorrectly for one set of cases.
4367 Description: Fix for potential infinite loop processing records in an llog.
4368 Details : If an llog record is corrupted/zeroed, it is possible to loop
4369 forever in llog_process(). Validate the llog record length
4370 and skip the remainder of the block on error.
4373 Frequency : occasional (liblustre only)
4375 Description: liblustre could not open files whose last component is a symlink
4376 Details : sysio_path_walk() would incorrectly pass the open intent to
4377 intermediate path components.
4380 Frequency : rare (liblustre only with non-standard tuning)
4381 Bugzilla : 7201 (7350)
4382 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4383 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4384 after it has been released, only do this for referenced locks
4386 Severity : enhancement
4388 Description: specifying an (invalid) directory default stripe_size of -1
4389 would reset the directory default striping
4390 Details : stripe_size -1 was used internally to signal directory stripe
4391 removal, now use "all default" to signal dir stripe removal
4392 as a directory striping of "all default" is not useful
4395 Frequency : common for large clusters running liblustre clients
4397 Description: doing an ls when liblustre clients are running is slow
4398 Details : sending a glimpse AST to a liblustre client waits for every AST
4399 to time out, as liblustre clients will not respond. Since they
4400 cannot cache data we refresh the OST lock LVB from disk instead.
4402 Severity : enhancement
4404 Description: doing an ls at the same time as file IO can be slow
4405 Details : enqueue and other "small" requests can be blocked behind many
4406 large IO requests. Create a new OST IO portal for non-IO
4407 requests so they can be processed faster.
4410 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4412 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4413 created file with mode 000
4414 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4415 allow an ftruncate() as a non-root user to a file with mode 000.
4416 Lustre now respects this flag to disable mode checks when
4417 truncating a file owned by the user
4420 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4423 Description: Revoking locks from clients that went dead or catatonic might take
4425 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4426 cancellation of such locks instant on servers without waiting for
4427 any reply from clients. Clients drops these locks when cancel
4428 notification from server is received without replying.
4431 Frequency : liblustre-only, when liblustre client dies or becomes busy
4433 Description: Doing ls on Linux clients can take a long time with active
4435 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4436 granting such locks to it in the first place if possible. Locks
4437 are taken by proxy on the OST during the read or write and
4438 dropped immediately afterward. Add connect flags handling, do
4439 not grant locks to liblustre clients for glimpse ASTs.
4441 Severity : enhancement
4443 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4444 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4445 sequential reads and behaves badly with applications doing random
4446 reads. Improve it by reading ahead at least read region, and
4447 avoiding excessive large RPC for small reads.
4449 Severity : enhancement
4451 Description: Creating more than 1000 files for a single job may cause a load
4452 imbalance on the OSTs if there are also a large number of OSTs.
4453 Details : qos_prep_create() uses an OST index reseed value that is an
4454 even multiple of the number of available OSTs so that if the
4455 reseed happens in the middle of the object allocation it will
4456 still utilize the OSTs as uniformly as possible.
4461 Description: OST or MDS may oops in ping_evictor_main()
4462 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4463 but doesn't restart at beginning of obd_exports_timed list
4466 Severity : enhancement
4468 Description: improve by-nid export eviction on the MDS and OST
4469 Details : allow multiple exports with the same NID to be evicted at one
4470 time without re-searching the exports list.
4473 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4475 Description: MDS may oops in groups_free()
4476 Details : in rare race conditions a newly allocated group_info struct is
4477 freed again, and this can be NULL. The 2.4 compatibility code
4478 for groups_free() checked for a NULL pointer, but 2.6 did not.
4481 Frequency : common for liblustre clients doing little filesystem IO
4482 Bugzilla : 9352, 7313
4483 Description: server may evict liblustre clients accessing contended locks
4484 Details : if a client is granted a lock or receives a completion AST
4485 with a blocking AST already set it would not reply to the AST
4486 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4489 Frequency : lfs setstripe, only systems with more than 160 OSTs
4491 Description: unable to set striping with a starting offset beyond OST 160
4492 Details : llapi_create_file() incorrectly limited the starting stripe
4493 index to the maximum single-file stripe count.
4496 Frequency : LDAP users only
4498 Description: lconf did not handle in-kernel recovery with LDAP properly
4499 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4501 Severity : enhancement
4503 Description: bind OST threads to NUMA nodes to improve performance
4504 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4505 node and do their allocations there to localize memory access
4507 Severity : enhancement
4509 Description: llmount can determine client NID directly from Myrinet (GM)
4510 Details : the client NID code from gmnalnid was moved directly into
4511 llmount, removing the need to use this or specifying the
4512 client NID explicitly when mounting GM clients with zeroconf
4515 Frequency : if client is started with down MDS
4517 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4518 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4519 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4520 ptlrpc_import_delayed_req() because we didn't check if the
4521 request was interrupted, and we also didn't break out of the
4522 event loop if there was no timeout
4527 Description: data loss during non-page-aligned writes to a single file from
4528 both multiple nodes and multiple threads on one node at same time
4529 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4530 inconsistency led to false short-reads, that were cached and later
4531 used by ->prepare_write() to fill in partially written page,
4532 leading to data loss.
4535 Frequency : always, if lconf --abort_recovery used
4537 Description: lconf --abort_recovery fails with 'Operation not supported'
4538 Details : lconf was attempting to abort recovery on the MDT device and not
4541 Severity : enhancement
4543 Description: remove cleanup logs
4544 Details : replace lconf-generated cleanup logs with lustre internal
4545 cleanup routines. Eliminates the need for client-cleanup and
4548 Severity : enhancement
4550 Description: add support for EAs (user and system) on lustre filesystems
4551 Details : it is now possible to store extended attributes in the Lustre
4552 client filesystem, and with the user_xattr mount option it
4553 is possible to allow users to store EAs on their files also
4555 Severity : enhancement
4557 Description: Add possibility (config option) to show minimal available OST free
4559 Details : When compiled with --enable-mindf configure option, statfs(2)
4560 (and so, df) will return least minimal free space available from
4561 all OSTs as amount of free space on FS, instead of summary of
4562 free spaces of all OSTs.
4564 Severity : enhancement
4566 Description: do not expand extent locks acquired on OST-side
4567 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4568 by server: they are not cached anyway.
4571 Frequency : when mmap is used/binaries executed from Lustre
4573 Description: Unmmap pages before throwing them away from read cache.
4574 Details : llap_shrink cache now attempts to unmap pages before discarding
4575 them (if unmapping failed - do not discard). SLES9 kernel has
4576 extra checks that trigger if this unmapping is not done first.
4581 Description: lconf didn't resolve symlinks before checking to see whether a
4582 given mountpoint was already in use
4585 Frequency : when migrating failover services
4586 Bugzilla : 6395, 9514
4587 Description: When migrating a subset of services from a node (e.g. failback
4588 from a failover service node) the remaining services would
4589 time out and evict clients.
4590 Details : lconf --force (implied by --failover) sets the global obd_timeout
4591 to 5 seconds in order to quickly disconnect, but this caused
4592 other RPCs to time out too quickly. Do not change the global
4593 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4595 Severity : enhancement
4596 Frequency : if MDS is started with down OST
4597 Bugzilla : 9439,5706
4598 Description: Allow startup/shutdown of an MDS without depending on the
4599 availability of the OSTs.
4600 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4601 Add appropriate locking and lov-osc refcounts for safe
4602 cleaning. Add osc abort_inflight calls in case the
4603 synchronize never started.
4606 Frequency : occasional (Cray XT3 only)
4608 Description: root not authorized to access files in CRAY_PORTALS environment
4609 Details : The client process capabilities were not honoured on the MDS in
4610 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4611 been accessed by an authorized user then root was able to access
4612 the file on the local client also. The root user capabilities
4613 are now allowed on the MDS, as this environment has secure UID.
4616 Frequency : occasional
4618 Description: ldiskfs "too long searching" message happens too often
4619 Details : A debugging message (otherwise harmless) prints too often on
4620 the OST console. This has been reduced to only happen when
4621 there are fragmentation problems on the filesystem.
4626 Description: Division by zero in statfs when all OSCs are inactive
4627 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4628 lov_statfs divides by value returned by lov_get_stripecnt().
4632 Bugzilla : 9489, 3273
4633 Description: First write from each client to each OST was only 4kB in size,
4634 to initialize client writeback cache, which caused sub-optimal
4635 RPCs and poor layout on disk for the first writen file.
4636 Details : Clients now request an initial cache grant at (re)connect time
4637 and so that they can start streaming writes to the cache right
4638 away and always do full-sized RPCs if there is enough data.
4639 If the OST is rebooted the client also re-establishes its grant
4640 so that client cached writes will be honoured under the grant.
4645 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4646 Details : Now I/O RPCs go to different portal number and (presumably) fast
4647 lock requests (and glimses) and other RPCs get their own service
4648 threads pool that should be able to service those RPCs
4651 Severity : enhancement
4653 Description: Ability to exchange lustre version between client and servers and
4654 issue warnings at client side if client is too old. Also for
4655 liblustre clients there is ability to refuse connection of too old
4657 Details : New 'version' field is added to connect data structure that is
4658 filled with version info. That info is later checked by server and
4662 Frequency : rare, liblustre only.
4663 Bugzilla : 9296, 9581
4664 Description: Two simultaneous writes from liblustre at offset within same page
4665 might proceed at the same time overwriting eachother with stale
4667 Details : I/O lock withing llu_file_prwv was released too early, before data
4668 actually was hitting the wire. Extended lock-holding time until
4669 server acknowledges receiving data.
4672 Frequency : extremely rare. Never observed in practice.
4674 Description: avoid generating lustre_handle cookie of 0.
4675 Details : class_handle_hash() generates handle cookies by incrementing
4676 global counter, and can hit 0 occasionaly (this is unlikely, but
4677 not impossible, because initial value of cookie counter is
4678 selected randonly). Value of 0 is used as a sentinel meaning
4679 "unassigned handle" --- avoid it. Also coalesce two critical
4680 sections in this function into one.
4682 Severity : enhancement
4684 Description: allow liblustre clients to delegate truncate locking to OST
4685 Details : To avoid overhead of locking, liblustre client instructs OST to
4686 take extent lock in ost_punch() on client's behalf. New connection
4687 flag is added to handle backward compatibility.
4689 Severity : enhancement
4690 Bugzilla : 4928, 7341, 9758
4691 Description: allow number of OST service threads to be specified
4692 Details : a module parameter allows the number of OST service threads
4693 to be specified via "options ost ost_num_threads={N}" in the
4694 OSS's /etc/modules.conf or /etc/modprobe.conf.
4698 Bugzilla : 6146, 9635, 9895
4699 Description: servers crash with bad pointer in target_handle_connect()
4700 Details : In rare cases when a client is reconnecting it was possible that
4701 the connection request was the last reference for that export.
4702 We would temporarily drop the export reference and get a new
4703 one, but this may have been the last reference and the export
4704 was just destroyed. Get new reference before dropping old one.
4706 Severity : enhancement
4707 Frequency : if client is started with failover MDS
4709 Description: Allow multiple MDS hostnames in the mount command
4710 Details : Try to read the configuration from all specified MDS
4711 hostnames during a client mount in case the "primary"
4714 Severity : enhancement
4716 Description: Stop sending data to evicted clients as soon as possible.
4717 Details : Check if the client we are about to send or are sending data to
4718 was evicted already. (Check is done every second of waiting,
4719 for which l_wait_event interface was extended to allow checking
4720 of exit condition at specified intervals).
4723 Frequency : rare, normally only when NFS exporting is done from client
4725 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4727 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4730 Frequency : always, if srand() is called before liblustre initialization
4732 Description: Liblustre uses system PRNG disturbing its usage by user application
4733 Details : Introduce internal to lustre fast and high-quality PRNG for
4734 lustre usage and make liblustre and some other places in generic
4735 lustre code to use it.
4737 Severity : enhancement
4738 Bugzilla : 9477, 9557, 9870
4739 Description: Verify that the MDS configuration logs are updated when xml is
4740 Details : Check if the .xml configuration logs are newer than the config
4741 logs stored on the MDS and report an error if this is the case.
4742 Request --write-conf, or allow starting with --old_conf.
4744 Severity : enhancement
4746 Description: Handle symlinks in the path when checking if Lustre is mounted.
4747 Details : Resolve intermediate symlinks when checking if a client has
4748 mounted a filesystem to avoid duplicate client mounts.
4753 Description: lconf can hit an error exception but still return success.
4754 Details : The lconf command catches the Command error exception at the top
4755 level script context and will exit with the associated exit
4756 status, but doesn't ensure that this exit status is non-zero.
4761 Description: failure of ptlrpc thread startup can cause oops
4762 Details : Starting a ptlrpc service thread can fail if there are a large
4763 number of threads or the server memory is very fragmented.
4764 Handle this without oopsing.
4767 Frequency : always, only if liblustre and non-default acceptor port was used
4769 Description: liblustre cannot connect to servers with non-default acceptor port
4770 Details : tcpnal_set_default_params() was not called and was therefore
4771 ignoring the environment varaible TCPNAL_PORT, as well as other
4772 TCPNAL_ environment variables
4777 Description: two objects could be created on the same OST for a single file
4778 Details : If an OST is down, in some cases it was possible to create two
4779 objects on a single OST for a single file. No problems other
4780 than potential performance impact and spurious error messages.
4784 Bugzilla : 5681, 9562
4785 Description: Client may oops in ll_unhash_aliases
4786 Details : Client dcache may become inconsistent in race condition.
4787 In some cases "getcwd" can fail if the current directory is
4793 Description: Inode refcounting problems in NFS export code
4794 Details : link_raw functions used to call d_instantiate without obtaining
4795 extra inode reference first.
4799 Bugzilla : 9942, 9903
4800 Description: Referencing freed requests leading to crash, memleaks with NFS.
4801 Details : We used to require that call to ll_revalidate_it was always
4802 followed by ll_lookup_it. Also with revalidate_special() it is
4803 possible to call ll_revalidate_it() twice for the same dentry
4804 even if first occurence returned success. This fix changes semantic
4805 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4806 reference on a request referred from the intent.
4807 ll_intent_release() then releases such a request.
4810 Frequency : rare, normally benchmark loads only
4812 Description: unlinked inodes were kept in memory on the client
4813 Details : If a client is repeatedly creating and unlinking files it
4814 can accumulate a lot of stale inodes in the inode slab cache.
4815 If there is no other client load running this can cause the
4816 client node to run out of memory. Instead flush old inodes
4817 from client cache that have the same inode number as a new inode.
4820 Frequency : SLES9 2.6.5 kernel and long filenames only
4821 Bugzilla : 9969, 10379
4822 Description: utime reports stale NFS file handle
4823 Details : SLES9 uses out-of-dentry names in some cases, which confused
4824 the lustre dentry revalidation. Change it to always use the
4828 Frequency : rare, unless heavy write-truncate concurrency is continuous
4829 Bugzilla : 4180, 6984, 7171, 9963, 9331
4830 Description: OST becomes very slow and/or deadlocked during object unlink
4831 Details : filter_destroy() was holding onto the parent directory lock
4832 while truncating+unlinking objects. For very large objects this
4833 may block other threads for a long time and slow overall OST
4834 responsiveness. It may also be possible to get a lock ordering
4835 deadlock in this case, or run out of journal credits because of
4836 the combined truncate+unlink. Solution is to do object truncate
4837 first in one transaction without parent lock, and then do the
4838 final unlink in a new transaction with the parent lock. This
4839 reduces the lock hold time dramatically.
4842 Frequency : rare, 2.4 kernels only
4844 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4845 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4846 before unmounting their filesystem in order to avoid deadlock.
4847 The kernel_locked() function in 2.4 kernels only checks whether
4848 the kernel lock is held, not whether it is this process that is
4849 holding it as 2.6 kernels do.
4854 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4855 Details : The client ptlrpc code may be trying to reconnect to a down
4856 server before a previous connection attempt has timed out.
4857 Increase the reconnect interval to be longer than the connection
4858 timeout interval to avoid sending duplicate connections to
4862 Frequency : echo_client brw_test command
4864 Description: fix echo_client to work with OST preallocated code
4865 Details : OST preallocation code (5137) didn't take echo_client IO path
4866 into account: echo_client calls filter methods outside of any
4867 OST thread and, hence, there is no per-thread preallocated
4868 pages and buffers to use. Solution: hijack pga pages for IO. As
4869 a byproduct, this avoids unnecessary data copying.
4873 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4874 Description: Client can oops in mdc_commit_close() after open replay
4875 Details : It was possible for the MDS to return an open request with no
4876 transaction number in mds_finish_transno() if the client was
4877 evicted, but without actually returning an error. Clients
4878 would later try to replay that open and may trip an assertion
4879 Simplify the client close codepath, and always return an error
4880 from the MDS in case the open is not successful.
4883 Frequency : rare, 2.6 OSTs only
4885 Description: OST may deadlock under high load on fragmented files
4886 Details : If there was a heavy load and highly-fragmented OST filesystems
4887 it was possible to have all the OST threads deadlock waiting on
4888 allocation of biovecs, because the biovecs were not released
4889 until the entire RPC IO was completed. Instead, release biovecs
4890 as soon as they are complete to ensure forward IO progress.
4892 Severity : enhancement
4894 Description: Support for specifying external journal device at mount
4895 Details : If an OST or MDS device is formatted with an external journal
4896 device, this device major/minor is stored in the ext3 superblock
4897 and may not be valid for failover. Allow detecting and
4898 specifying the external journal at mount time.
4903 Description: Mounting an MDS with pending unlinked files may cause oops
4904 Details : target_finish_recovery() calls mds_postrecov() which returned
4905 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4906 considers this an error and immediately begins cleaning up the
4907 lov, just after starting the mds_lov process
4909 Severity : enhancement
4911 Description: Implement 'lfs df' to report actual free space on per-OST basis
4912 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4913 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4914 usage of objects; '-h' to report in human readable format.
4916 ------------------------------------------------------------------------------
4918 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4925 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4926 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4927 at startup during orphan unlinking. Instead, skip buddy update
4928 before setup, as it will be regenerated after recovery is complete.
4933 Description: If an OST is inactive, its locks might reference stale inodes.
4934 Details : lov_change_cbdata() must iterate over all namespaces, even if
4935 they are inactive to clear inode references from the lock.
4937 Severity : enhancement
4938 Frequency : occasional, if non-standard max_dirty_mb used
4940 Description: Client will block write RPCs if not enough grant
4941 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4942 then the client will block writes while waiting for another RPC
4943 to complete instead of consuming its dirty limit. With change
4944 we get improved performance when max_dirty_mb is small.
4946 Severity : enhancement
4947 Bugzilla : 3389, 6253
4948 Description: Add support for supplementary groups on the MDS.
4949 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4950 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4951 lookups for user supplementary groups into a cache.
4955 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4956 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4957 for liblustre O_EXCL is enforced by MDS.
4961 Description: readdir never returns NULL in liblustre.
4962 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4963 dentry in struct dirent.
4967 Frequency : liblustre only, depends on application IO pattern
4968 Description: liblustre clients evicted if not contacting servers
4969 Details : Don't put liblustre clients into the ping_evictor list, so
4970 they will not be evicted by the pinger ever.
4972 Severity : enhancement
4974 Description: Add ability to evict clients by NID from MDS.
4975 Details : By echoing "nid:$NID" string into
4976 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4977 $NID would be instantly evicted from this MDS and from all active
4978 OSTs connected to it.
4982 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4983 Details : lookup_it_finish() used to query file size from OSTs that was not
4988 Description: service threads change working directory to that of init
4989 Details : Starting lustre service threads may pin the working directory
4990 of the parent thread, making that filesystem busy. Threads
4991 now change to the working directory of init to avoid this.
4995 Frequency : during shutdown only
4996 Description: shutdown with a failed MDS or OST can cause unmount to hang
4997 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5002 Frequency : 2.6 only
5003 Description: chmod/chown may include an extra supplementary group
5004 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5005 supplementary group and if none is specified this is used.
5008 Bugzilla : 5479 (6816)
5010 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5011 Details : If lookup is in progress on a file that is unlinked we might try
5012 to revalidate the inode and fail in revalidate after lookup is
5013 complete and ll_file_open() enqueues the open again but
5014 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5017 Frequency : always, if lconf --abort_recovery used
5019 Description: lconf --abort_recovery fails with 'Operation not supported'
5020 Details : lconf was attempting to abort recovery on the MDT device and not
5023 ------------------------------------------------------------------------------
5025 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5030 Frequency : rare (only unsupported configurations with a node running as an
5032 Bugzilla : 6514, 5137
5033 Description: Mounting a Lustre file system on a node running as an OST could
5035 Details : OSTs now preallocates memory needed to write out data at
5036 startup, instead of when needed, to avoid having to
5037 allocate memory in possibly low memory situations.
5038 Specifically, if the file system is mounted on on OST,
5039 memory pressure could force it to try to write out data,
5040 which it needed to allocate memory to do. Due to the low
5041 memory, it would be unable to do so and the node would
5042 become unresponsive.
5044 Severity : enhancement
5046 Description: Addition of lconf --service command line option
5047 Details : lconf now accepts a '--service <arg>' option, which is
5048 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5050 Severity : enhancement
5052 Description: Failover mode is now the default for OSTs.
5053 Details : By default, OSTs will now run in failover mode. To return to
5054 the old behaviour, add '--failout' to the lmc line for OSTs.
5056 Severity : enhancement
5058 Description: Health checks are now provided for MDS and OSTs
5059 Details : Additional detailed health check information on MSD and OSTs
5060 is now provided through the procfs health_check value.
5063 Frequency : occasional, depends on IO load
5065 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5066 after numerous create/delete cycles
5067 Details : The ext3 inode allocation policy would not allocate new inodes
5068 very well on the OSTs because there are no new directories
5069 being created. Instead we look for groups with free space if
5070 the parent directories are nearly full.
5075 Description: Network or server problems during mount may cause partially
5076 mounted clients instead of returning an error.
5077 Details : The config llog parsing code may overwrite the error return
5078 code during mount error handling, returning success instead
5084 Description: MDS can fail to allocate large reply buffers
5085 Details : After long uptimes the MDS can fail to allocate large reply
5086 buffers (e.g. zconf client mount config records) due to memory
5087 fragmentation or consumption by the buffer cache. Preallocate
5088 some large reply buffers so that these replies can be sent even
5089 under memory pressure.
5093 Frequency : rare (liblustre)
5094 Description: fsx running with liblustre complained that using truncate() to
5095 extend the file doesn't work. This patch corrects that issue.
5096 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5097 ATTR_SIZE and lsm use in llu_setattr_raw.
5101 Frequency : rare, only 2.6 kernels
5102 Description: Unusual file access patterns on the MDS may result in inode
5103 data being lost in very rare circumstances.
5104 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5105 code in the 2.6 kernel caused the inode and/or EA data not to
5106 be read from disk, causing single-file corruption.
5110 Frequency : rare, only 2.6 filesystems using extents
5111 Description: Heavy concurrent write and delete load may cause data corruption.
5112 Details : It was possible under high-load situations to have an extent
5113 metadata block in the block device cache from a just-unlinked
5114 file overwrite a newly-allocated data block. We now unmap any
5115 metadata buffers that alias just-allocated data blocks.
5119 Frequency : filesystems with default stripe_count larger than 77
5120 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5121 Details : lconf specifies an inode size of 4096 bytes when the default
5122 stripe_count is larger than 77. This conflicts with the default
5123 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5124 this case to avoid pinning too much memory for large EAs.
5126 ------------------------------------------------------------------------------
5128 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5133 Frequency : rare (extremely heavy IO load with hundreds of clients)
5135 Description: Client is evicted, gets IO error writing to file
5136 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5137 caused clients to be evicted for AST timeouts. The fixes in
5138 bug 5192 mean we no longer need to have such short AST timeouts
5139 so ldlm_timeout has been increased.
5142 Frequency : occasional during --force or --failover shutdown under load
5143 Bugzilla : 5949, 4834
5144 Description: Server oops/LBUG if stopped with --force or --failover under load
5145 Details : a collection of import/export refcount and cleanup ordering
5146 issues fixed for safer force cleanup
5149 Frequency : only filesystems larger than 120 OSTs
5150 Bugzilla : 5990, 6223
5151 Description: lfs getstripe would oops on a very large filesystem
5152 Details : lov_getconfig used kfree on vmalloc'd memory
5155 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5156 Bugzilla : 6242, 6243
5157 Description: reading from files that had been truncated to a non-zero size
5158 but never opened returned no data
5159 Details : ll_file_read() reads zeros from no-object files to EOF
5164 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5165 Details : There's little harm in aborting MDS/OSS recovery and letting it
5166 try again, so I removed the LASSERT and return an error instead.
5168 Severity : enhancement
5170 Description: New debugging infrastructure for tracking down data corruption
5171 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5172 (b) cover more of the client-side code path, and (c) try to narrow
5173 down where problems occurred
5177 Bugzilla : 3819, 4364, 4397, 6313
5178 Description: Racing close and eviction MDS could cause assertion in mds_close
5179 Details : It was possible to get multiple mfd references during close and
5180 client eviction, leading to one thread referencing a freed mfd.
5182 Severity: : enhancement
5183 Bugzilla : 3262, 6359
5184 Description: Attempts to reconnect to servers are now more aggressive.
5185 Details : This builds on the enhanced upcall-less recovery that was added
5186 in 1.4.2. When trying to reconnect to servers, clients will
5187 now try each server in the failover group every 10 seconds. By
5188 default, clients would previously try one server every 25 seconds.
5193 Description: After recovery, certain operations trigger a failed
5194 assertion on a client.
5195 Details : Failing over an mds, using lconf -d --failover, while a
5196 client was doing a readdir() call would cause the client to
5197 LBUG after recovery completed and the readdir() was resent.
5199 Severity : enhancement
5201 Description: Default groups are now added by lconf
5202 Details : You can now run lconf --group <servicename> without having to
5203 manually add groups with lmc.
5206 Frequency : occasional
5208 Description: Nodes with an elan id of 0 trigger a failed assertion
5211 Frequency : always when accessing e.g. tty/console device nodes
5213 Description: tty and some other devices nodes cannot be used on lustre
5214 Details : file's private_data field is used by device data and lustre
5215 values in there got lost. New field was added to struct file to
5216 store fs-specific private data.
5219 Frequency : when exporting Lustre via NFS
5221 Description: NFSD failed occasionally when looking up a path component
5222 Details : NFSD is looking up ".." which was broken in ext3 directories
5223 that had grown large enough to become hashed.
5226 Frequency : Clusters with multiple interfaces not on the same subnet
5228 Description: Nodes will repeatedly try to reconnect to an interface which it
5229 cannot reach and report an error to the log.
5230 Details : Extra peer list entries will be created by lconf with some peers
5231 unreachable. lconf now validates the peer before adding it.
5234 Frequency : Only if a default stripe is set on the filesystem root.
5236 Description: Setting a default stripe on the filesystem root prevented the
5237 filesystem from being remounted.
5238 Details : The client was sending extra request flags in the root getattr
5239 request and did not allocate a reply buffer for the dir EA.
5242 Frequency : occasional, higher if lots of files are accessed by one client
5243 Bugzilla : 6159, 6097
5244 Description: Client trips assertion regarding lsm mismatch/magic
5245 Details : While revalidating inodes the VFS looks up inodes with ifind()
5246 and in rare cases can find an inode that is being freed.
5247 The ll_test_inode() code will free the lsm during ifind()
5248 when it finds an existing inode and then the VFS later attaches
5249 this free lsm to a new inode.
5253 Bugzilla : 6422, 7030
5254 Description: MDS deadlock between mkdir and client eviction
5255 Details : Creating a new file via mkdir or mknod (starting a transaction
5256 and getting the ns lock) can deadlock with client eviction
5257 (gets ns lock and trying to finish a synchronous transaction).
5260 Frequency : occasional
5261 Description: While starting a server, the fsfilt_ext3 module could not be
5263 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5264 kernels. Previously, lconf would still use the ext3 name
5265 when trying to load modules. Now, it will correctly use
5266 ext3 on 2.4 and ldiskfs on 2.6.
5268 Severity : enhancement
5269 Description: The default stripe count has been changed to 1
5270 Details : The interpretation of the default stripe count (0, to lfs
5271 or lmc) has been changed to mean striping across a single
5272 OST, rather than all available. For general usage we have
5273 found a stripe count of 1 or 2 works best.
5275 Severity : enhancement
5276 Description: Add support for compiling against Cray portals.
5277 Details : Conditional compiling for some areas that are different
5281 Frequency : occasional
5282 Bugzilla : 6409, 6834
5283 Description: Creating files with an explicit stripe count may lead to
5284 a failed assertion on the MDS
5285 Details : If some OSTs are full or unavailable, creating files may
5286 trigger a failed assertion on the MDS. Now, Lustre will
5287 try to use other servers or return an error to the
5291 Frequency : occasional
5293 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5294 caused lock timeout during readahead (since 1.4.2).
5295 Details : Processes doing readahead might match a lock that hasn't been
5296 granted yet if there are overlapping and conflicting lock
5297 requests. The readahead process waits on ungranted lock
5298 (original lock is CBPENDING), while OST waits for that process
5299 to cancel CBPENDING read lock and eventually evicts client.
5301 Severity : enhancement
5303 Description: Initial enabling of flock support for clients
5304 Details : Implements fcntl advisory locking and file status functions.
5305 This feature is provided as an optional mount flag (default
5306 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5307 locking are implemented yet, and those that are are not guaranteed
5308 to be completely correct in production environments.
5309 mount -t lustre -o [flock|noflock] ...
5312 Frequency : occasional
5314 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5315 assertion in the ext3 JBD (journaling) layer.
5316 Details : The b_committed_data struct is protected by the big kernel lock
5317 in 2.4 kernels, serializing journal_commit_transaction() and
5318 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5319 access to this struct. In 2.6 kernels there is finer grained
5320 locking to improve SMP performance of the JBD layer.
5324 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5329 Description: merge and cleanup kernel patches.
5332 -----------------------------------------------------------------------------
5334 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5336 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5337 versions of Lustre. Please update all servers and clients to
5338 version 1.4.2 or later at the same time. You must also run
5339 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5340 to update the configuration logs.
5342 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5343 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5344 - swab llog records of type '0' so we get proper header size/idx (5861)
5345 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5346 - fix rename of one directory over another leaking an inode (5953)
5347 - avoid SetPageDirty on 2.6 (5981)
5348 - don't re-add just-being-destroyed locks to the waiting list (5653)
5349 - when creating new directories, inherit the parent's custom
5350 striping settings if present parent (3048)
5351 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5352 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5353 - don't hold client locks on temporary worklist from l_lru (5666)
5354 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5355 - automatically evict dead clients (5921)
5356 - Update file size properly in create+truncate+fstat case (6196)
5357 - Do not unhash mountpoint dentries, do not allow removal of
5359 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5360 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5361 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5362 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5363 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5364 - Correctly handle reads of files with no objects (6243)
5365 - lctl recover will also mark a device active if deactivate used (5933)
5367 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5368 - allow --write-conf on an MDS with different nettype than client (5619)
5369 - don't write config llogs to MDS for mounts not from that MDS (5617)
5370 - lconf should create multiple TCP connections from a client (5201)
5371 - init scripts are now turned off by default; run chkconfig --on
5372 lustre and chkconfig --on lustrefs to use them
5373 - upcalls are no longer needed for clients to recover to failover
5375 - add --abort-recovery option to lconf to abort recovery on device
5377 - add support for an arbitrary number of OSTs (3026)
5378 - Quota support protocol changes.
5379 - forward compatibility changes to wire structs (6007)
5380 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5381 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5382 - improved "lustre status" script
5383 - initialize blocksize for non-regular files (6062)
5384 - added --disable-server and --disable-client configure options (5782)
5385 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5386 - Vanilla 2.4.29 support
5387 - increase maximum number of obd devices to 520 (6242)
5388 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5389 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5390 SLES 9 2.6.5 kernels
5391 - limit stripes per file to 160 (the maximum EA size) (6093)
5393 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5396 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5397 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5398 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5399 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5400 - small ext3 extents cleanups and fixes (5733)
5401 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5402 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5403 - avoid needless client->OST connect, fix handle mismatch (5317)
5404 - fix DLM error path that led to out-of-sync client, long delays (5779)
5405 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5406 - fix several locking issues related to i_size (5492,5624,5654,5672)
5407 - don't move pending lock onto export if it is already evicted (5683)
5408 - fix kernel oops when creating .foo in unlinked directory (5548)
5409 - fix deadlock in obdfilter statistics vs. object create (5811)
5410 - use time_{before,after} to avoid timer jiffies wrap (5882)
5411 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5412 - Do not leak request if server was not able to process it (5154)
5413 - If mds_open unable to find parent dir, make that negative lookup(5154)
5414 - don't create new directories with extent-mapping (5909, 5936)
5416 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5417 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5418 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5419 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5420 - support for RHEL4 kernel on i686 (5773)
5421 - provide error messages when incompatible logs are encountered (5898)
5423 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5424 * version 1.4.0.10 (1.4.1 release candidate 1)
5426 - don't keep a lock reference when lock is not granted (4238)
5427 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5428 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5429 - drop import inflight refcount on signal_completed_replay error (5255)
5430 - unlock page after async write error during send (3677)
5431 - handle missing objects in filter_preprw_read properly (5265)
5432 - no transno return for symlink open, don't save no-trasno open (3440)
5433 - don't try to complete elan receive that already failed (4012)
5434 - free RPC server reply state on error (5406)
5435 - clean up thread from ptlrpc_start_thread() on error (5160)
5436 - readahead could read extra page into cache that wasn't ejected (5388)
5437 - prevent races in class_attach/setup/cleanup/detach (5260)
5438 - don't dereference de->d_inode after l_dput of de (5458)
5439 - use "int" for stripe value returned from lock_to_stripe (5544)
5440 - mballoc allocation and error-checking fixes in 2.6 (5504)
5441 - block device patches to fix I/O request sizes in 2.6 (5482)
5442 - look up hostnames for IB nals (5602)
5443 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5444 - don't start multiple acceptors for the same port (5277)
5445 - fix incorrect LASSERT in mds_getattr_name (5635)
5446 - export a proc file for general "ping" checking (5628)
5447 - fix "lfs check" to not block when the MDS is down (5628)
5449 - service request history (4965)
5450 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5451 - create an "evict_client" /proc entry on OSTs, like the MDS has
5452 - fix mount usage message, return errors per mount(8) (5168)
5453 - change grep [] to grep "[]" in tests so they work in more UMLs
5454 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5455 - remove ancient LOV_MAGIC_V0 EA support (5047)
5456 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5457 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5458 - allow readahead allocations to fail when low on memory (5383)
5459 - mmap locking landed again, after considerable improvement (2828)
5460 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5462 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5465 - send OST transaction number in read/write reply to free req (4966)
5466 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5467 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5468 - the watchdog thread now runs as interruptible (5246)
5469 - flock/lockf fixes (but it's still disabled, pending 5135)
5470 - don't use EXT3 constants in llite code (5094)
5471 - memory shortage at startup could cause assertion (5176)
5473 - reorganization of lov code
5474 - single portals codebase
5476 - add extents/mballoc support (5025)
5477 - direct I/O reads in the obdfilter (4048)
5478 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5480 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5483 - send OST transaction number in read/write reply to free req (4966)
5484 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5485 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5486 - the watchdog thread now runs as interruptible (5246)
5487 - handle missing objects in filter_preprw_read properly (5265)
5488 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5489 - drop import inflight refcount on signal_completed_replay error (5255)
5490 - unlock page after async write error during send (3677)
5491 - return original error code on reconstructed replies (3761)
5492 - no transno return for symlink open, don't save no-trasno open (3440)
5494 - add pid to ldlm debugging output (4922)
5495 - bump the watchdog timeouts -- we can't handle 30sec yet
5496 - extra debugging for orphan dentry/inode bug (5259)
5498 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5501 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5502 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5503 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5504 - fix NULL dereference in /proc/sys/portals/routes (4827)
5505 - allow failed mdc_close() operations to be interrupted (4561)
5506 - stop precreate on OST before MDS would time out on it (4778)
5507 - don't send partial-page writes before EOF from client (4410)
5508 - discard client grant for sub-page writes on large-page clients (4520)
5509 - don't free dentries not owned by NFS code, check generation (4806)
5510 - fix lsm leak if mds_create_objects() fails (4801)
5511 - limit debug_daemon file size, always print CERROR messages (4789)
5512 - use transno after validating reply (3892)
5513 - process timed out requests if import state changes (3754)
5514 - update mtime on OST during writes, return in glimpse (4829)
5515 - add mkfsoptions to LDAP (4679)
5516 - use ->max_readahead method instead of zapping global ra (5039)
5517 - don't interrupt __l_wait_event() during strace
5519 - add software watchdogs to catch hung threads quickly (4941)
5520 - make lustrefs init script start after nfs is mounted
5521 - fix CWARN/ERROR duplication (4930)
5522 - return async write errors to application if possible (2248)
5523 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5524 - print NAL number in %x format (4645)
5525 - update barely-supported suse-2.4.21-171 series (4842)
5526 - support for sles 9 %post scripts
5527 - support for building 2.6 kernel-source packages
5528 - support for sles km_* packages
5530 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5533 - ignore -ENOENT errors in osc_destroy (3639)
5534 - notify osc create thread that OSC is being cleaned up (4600)
5535 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5536 - reconstruct ost_handle() like mds_handle() (4657)
5537 - create a new thread to do import eviction to avoid deadlock (3969)
5538 - let lconf resolve symlinked-to devices (4629)
5539 - don't unlink "objects" from directory with default EA (4554)
5540 - hold socknal file ref over connect in case target is down (4394)
5541 - allow more than 32000 subdirectories in a single directory (3244)
5542 - fix blocks count for O_DIRECT writes (3751)
5543 - OST returns ENOSPC from object create when no space left (4539)
5544 - don't send truncate RPC if file size isn't changing (4410)
5545 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5546 - bind to privileged port in socknal and tcpnal (3689)
5548 - rate limit CERROR/CWARN console message to avoid overload (4519)
5549 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5550 - basic mmap support (3918)
5551 - kernel patch series update from b1_4 (4711)
5553 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5556 - avoid crash during MDS cleanup with OST shut down (2775)
5557 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5558 - don't use bad inodes on the MDS (3744)
5559 - dynamic object preallocation to improve recovery speed (4236)
5560 - don't hold spinlock over lock dumping or change debug flags (4401)
5561 - don't zero obd_dev when it is force cleaned (3651)
5562 - print grants to console if they go negative (4431)
5563 - "lctl deactivate" will stop automatic recovery attempts (3406)
5564 - look for existing locks in ldlm_handle_enqueue() (3764)
5565 - don't resolve lock handle twice in recovery avoiding race (4401)
5566 - revalidate should check working dir is a directory (4134)
5568 - don't always mark "slow" obdfilter messages as errors (4418)
5570 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5573 - don't close LustreDB during write_conf until it is done (3860)
5574 - fix typo in lconf for_each_profile (3821)
5575 - allow dumping logs from multiple threads at one time (3820)
5576 - don't allow multiple threads in OSC recovery (3812)
5577 - fix debug_size parameters (3864)
5578 - fix mds_postrecov to initialize import for llog ctxt (3121)
5579 - replace config semaphore with spinlock (3306)
5580 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5581 - don't allow enqueue to complete on a destroyed export (3822)
5582 - down write_lock before checking llog header bitmap (3825)
5583 - recover from lock replay timeout (3764)
5584 - up llog sem before sending rpc (3652)
5585 - reduce ns lock hold times when setting kms (3267)
5586 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5587 - fix NULL deref and obd_dev leak on setup error (3312)
5588 - replace some LBUG about llog ops with error handling (3841)
5589 - don't match INVALID dentries from d_lookup and spin (3784)
5590 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5591 - fix invalid assertion in ptlrpc_set_wait (3880)
5593 - add libwrap support for the TCP acceptor (3996)
5594 - add /proc/sys/portals/routes for non-root route listing (3994)
5595 - allow setting MDS UUID in .xml (2580)
5596 - print the stack of a process that LBUGs (4228)
5598 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5601 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5602 - make sure to unset replay flag from failed open requests (3440)
5603 - if default stripe count is 0, use OST count for inode size (3636)
5604 - update parent mtime/ctime on client for create/unlink (2611)
5605 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5606 - free recovery state on server during a forced cleanup (3571)
5607 - unregister_reply for resent reqs (3063)
5608 - loop back devices mounting and status check on 2.6 (3563)
5609 - fix resource-creation race that can provoke i_size == 0 (3513)
5610 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5611 - more debugging for page-accounting assertion (3746)
5612 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5613 - avoid deadlock after precreation failure (3758)
5614 - fix race and lock order deadlock in orphan handling (3450, 3750)
5615 - add validity checks when grabbing inodes from l_ast_data (3599)
5617 - add /proc/.../recovery_status to obdfilter (3428)
5618 - lightweight CDEBUG infrastructure, debug daemon (3668)
5619 - change default OSC RPC parameters to be better on small clusters
5620 - turn off OST read cache for files smaller than 32MB
5621 - install man pages and include them in rpms (3100)
5622 - add new init script for (un)mounting lustre filesystems (2593)
5623 - run chkconfig in %post for init scripts (3701)
5624 - drop scimac NAL (unmaintained)
5626 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5629 - clean kiobufs before and after use (3485)
5630 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5631 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5632 - change init script to fail more clearly if not run as root (1528)
5633 - allow clients to reconnect during replay (1742)
5634 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5635 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5636 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5637 - fix cond_resched (3554)
5638 - only evict unfinished clients after recovery (3515)
5639 - allow bulk resend, prevent data loss (3570)
5640 - dynamic ptlrpc request buffer allocation (2102)
5641 - don't allow unlinking open directory if it isn't empty (2904)
5642 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5643 - remove extraneous obd dereference causing LASSERT failure (3334)
5644 - don't use get_cycles() when creating temp. files on the mds (3156)
5645 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5646 - handle EEXIST for set-stripe, set proper directory name (3336)
5648 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5649 - fix ksocknal_fmb_callback() error messages (2918)
5651 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5654 - don't copy lvb into (possibly NULL) reply on error (2983)
5655 - don't deref dentry after dput, don't free lvb on error (2922)
5656 - use the kms to determine writeback rpc length (2947)
5657 - increment oti_logcookies when osc is inactive (2948)
5658 - update client's i_blocks count via lvb messages (2543)
5659 - handle intent open/close of special files properly (1557)
5660 - mount MDS with errors=remount-ro, like obdfilter (2009)
5661 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5662 - don't use cancelling-locks' kms values (2947)
5663 - use highest lock extent for kms, not last one (2925)
5664 - don't dereference ERR_PTR() dentry in error handling path (3107)
5665 - fix thread race in portals_debug_dumplog() (3122)
5666 - create lprocfs device entries at setup instead of at attach (1519)
5667 - common AST error handler, don't evict client on completion race (3145)
5668 - zero nameidata in detach_mnt in 2.6 (3118)
5669 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5670 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5671 - initialize RPC timeout timer earlier for 2.6 (3219)
5672 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5673 - print nal/nid for unknown nid (3258)
5674 - additional checks for oscc recovery before doing precreate (3284)
5675 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5676 - don't crash in mdc_close for bad permissions on open (3285)
5677 - zero i_rdev for non-device files (3147)
5678 - clear page->private before handing to FS, better assertion (3119)
5679 - tune the read pipeline (3236)
5680 - fix incorrect decref of invalidated dentry (2350)
5681 - provide read-ahead stats and refine rpc in flight stats (3328)
5682 - don't hold journal transaction open across create RPC (3313)
5683 - update atime on MDS at close time (3265)
5684 - close LDAP connection when recovering to avoid server load (3315)
5685 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5686 - don't leak open file on MDS after open resend (3325)
5687 - serialize filter_precreate and filter_destroy_precreated (3329)
5688 - loop device shouldn't call sync_dev() for nul device (3092)
5689 - clear page cache after eviction (2766)
5690 - resynchronize MDS->OST in background (2824)
5691 - refuse to mount the same filesystem twice on same mountpoint (3394)
5692 - allow llmount to create routes for mounting behind routers (3320)
5693 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5694 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5695 - fix rare problem with rename on htree directories (3417)
5697 - allow default OST striping configuration per directory (1414)
5698 - fix compilation for qswnal for 2.6 kernels (3125)
5699 - increase maximum number of MDS request buffers for large systems
5700 - change liblustreapi to be useful for external progs like lfsck (3098)
5701 - increase local configuration timeout for slow disks (3353)
5702 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5704 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5707 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5708 - don't overwrite extent policy data in reply if lock was blocked (2901)
5709 - drop filter export grants atomically with removal from device (2663)
5710 - del obd_self_export from work_list in class_disconnect_exports (2908)
5711 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5712 - swab reply message in mdc_close, other PPC fixes (2464)
5713 - fix destroying of named logs (2325)
5714 - overwrite old logs when running lconf --write_conf (2264)
5715 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5716 - fix race in target_handle_connect (2898)
5717 - mds_reint_create() should take same inode create lock (2926)
5718 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5719 - don't close files for self_export to avoid uninitialized obd (2936)
5720 - allow MDS with the same name as client node (2939)
5721 - hold dentry reference for closed log files for unlink (2325)
5722 - reserve space for all logs during transactions (2059)
5723 - don't evict page beyond end of stripe extent (2925)
5724 - don't oops on a deleted current working directory (2399)
5725 - handle hard links to targets without a parent properly (2517)
5726 - don't dereference NULL lock when racing during eviction (2867)
5727 - don't grow lock extents when lots of conflicting locks (2919)
5729 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5732 - account for cache space usage on clients to avoid data loss (974)
5733 - lfsck support in lustre kernel code (2349)
5734 - reduce journal credits needed for BRW writes (2370)
5735 - orphan handling to avoid losing space on client/server crashes
5736 - ptlrpcd can be blocked, stopping ALL progress (2477)
5737 - use lock value blocks to assist in proper KMS, faster stat (1021)
5738 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5739 - recovery for initial connections (2355)
5740 - fixes for mds_cleanup_orphans (1934)
5741 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5742 - block all file creations until orphan recovery completes (1901)
5743 - client remove rq_connection from request struct (2423)
5744 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5745 - recovery timer race (2670)
5746 - mdc_close recovey bug (2532)
5747 - ptlrpc cleanup bug (2710)
5748 - mds timeout on local locks (2588)
5749 - namespace lock held during RPCs (2431)
5750 - handle interrupted sync write properly (2503)
5751 - don't try to handle a message that hasn't been replied to (2699)
5752 - client assert failure during cleanup after abort recovery (2701)
5753 - leak mdc device after failed mount (2712)
5754 - ptlrpc_check_set allows timedout requests to complete (2714)
5755 - wait for inflight reqs when ptlrpcd finishes (2710)
5756 - make sure unregistered services are removed from the srv_list
5757 - reset bulk XID's when resending them (caught by 1138 test)
5758 - unregister_bulk after timeout
5759 - fix lconf error (2694)
5760 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5761 - readahead locks pages, leaves pending causing memory pressure (2673)
5762 - increase OST request buffers to 4096 on large machines (2729)
5763 - fix up permission of existing directories in simple_mkdir (2661)
5764 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5765 - don't assign transno to errored transactions (2742)
5766 - don't delete objects on OST if given a bogus objid from MDS (2751)
5767 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5768 - if rq_no_resend, then timeout request after recovery (2432)
5769 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5770 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5771 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5772 - MDS recovery shouldn't delete objects that it has given out (2730)
5773 - if enqueue arrives after completion, don't clobber LVB (2819)
5774 - don't unlock pages twice when trigger_group_io returns error (2814)
5775 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5776 - don't write pages to disk if there was an error (1450)
5777 - don't ping imports that have recovery disabled (2676)
5778 - take buffered bytes into account when balancing socknal conn (2817)
5779 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5780 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5781 - remove little-endian swabbing of llog records (1987)
5782 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5783 - retry reposting request buffers if they fail (1191)
5784 - grow extent at grant time to avoid granting a revoked lock (2809)
5785 - lock revoke doesn't evict page if covered by a second lock (2765)
5786 - disable VM readahead to avoid reading outside lock extents (2805)
5788 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5789 - updated kernel patches for hp-2.4.20 kernel (2681)
5791 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5794 - fix truncated write corruption (2366)
5795 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5797 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5798 - ensure len(uuid) < 37 in lmc (1171)
5799 - fix ia64 OOPS in llog_test (2255)
5800 - zero end of page at obdfilter for partial page writes (2648)
5801 - don't leave stale dentries around after renames (bug 2428)
5802 - fix timeouts when evicting a client with a single lock held (2642)
5803 - set deadline for the initial HELLO message to drain (2634)
5804 - print out dotted-quad IP addresses in the socknal (2302)
5806 - additional debugging for MDS client eviction problem (2443)
5807 - fix mkfsoptions support for osts (2603, 2604)
5809 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5812 - add series for the vanilla 2.6.0 kernel
5813 - add series for the vanilla 2.4.24 kernel
5814 - add series for a cray x86/64 UL kernel drop
5815 - fix xattr patches for the vanilla 2.4.19 series
5817 - generate true UUIDs in lmc (1171)
5818 - have portals stack dumping break in UML (2466)
5819 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5820 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5821 - wake callback waiting threads on client eviction (2460)
5822 - Add --ptldebug and --subsystem to lmc (1719)
5823 - update assertion to allow safe interrupt allocation
5824 - set rq_no_resend for cancel requests (2432)
5825 - recalculate ptlrpcd timeout after resend (2494)
5826 - call vfs_rmdir when removing pending directories (2368)
5827 - fix renaming a file to itself (2429)
5828 - lmc creates a default one-stripe lov (2454)
5829 - expand procfs space to handle large clusters (2326)
5830 - increase UML stack to avoid overflow
5831 - update lconf's list of debug and subsystem masks
5832 - fix lfs find --obd (2510)
5833 - /proc tunable for disabling filter read caching (2591)
5834 - stop rpm packages from altering slapd.conf (2301)
5835 - disable nagle in the socknal under 0conf (2578)
5836 - choose mds inode size based on stripe count (2572)
5837 - fix kernel-source rpm problems (2516)
5839 - add --disable-doc to avoid pdf generation (2421)
5840 - update documentation, tests, type-os, comments
5841 - avoid format warnings on ia64
5842 - remove the TOE NAL
5843 - tiny code cleanups by removing unused fields
5845 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5848 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5849 - remove the most blatant lies from BUILDING file (2371)
5850 - change default debug level to reasonable production setting
5851 - reduce client side cache size to reduce cache flush time
5852 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5853 - make TCP zerocopy and pinger support enabled by default (2476)
5854 - sync writes completed after process exits caused crashes (2319)
5855 - maintain correct mount count on the MDS (2356)
5856 - backout 1557, because 2316 wasn't really fixed
5857 - better file I/O statistics gathering in /proc
5858 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5859 - another kernel patch to fix zero-copy TCP function export
5860 - don't take duplicate lock when processing re-sent getattr (2420)
5861 - lctl uses obd_self_export instead of creating new conn (2353)
5862 - MDS/OST recovery case which requires object creation asserted (2425)
5863 - move lfs from /usr/sbin to /usr/bin in packages
5864 - fix race between mds_client_add and mds_client_free (2417)
5865 - use kmalloc instead of slabs in portals (2430)
5866 - don't create duplicate records when a failover MDS is present (2442)
5867 - remove unnecessary mount age check (2332)
5868 - don't remove directory inodes from locks prematurely (2451)
5869 - don't break if MDS service name is the same as hostname (2103)
5870 - fix races in client write RPC generation when cache full (2482)
5872 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5875 - remove now-unused request->rq_obd (278)
5876 - if an allocation fails, print out how much memory we've used (1933)
5877 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5878 - add the "configurable stack size" patch to most series files (1256)
5879 - ability to write large log records, for 100+ OST configs (2306)
5880 - fix NULL deref when filter_prep fails (2314)
5881 - fix operator precedence error in filter_sync
5882 - dynamic allocation of socknal TX descriptors (2315)
5883 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5884 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5885 - ensure that CWARN messages in Portals always get to the syslog
5886 - __init/__exit are not for prototype decls (ldlm_init/exit)
5887 - x86-64 compile warning fixes
5888 - fix gateway LMC keyword conflict (2318)
5889 - fix MDS lock inversions in getattr/reint paths (1844)
5890 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5891 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5892 - DLM race condition prevented some lock evictions (2328)
5893 - ENOMEM detection and retry on socknal sends (2230)
5894 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5895 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5896 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5897 - better error messages when a client is rejected during recovery (1505)
5898 - avoid cancelling locks which were never granted, after failure (2330)
5899 - fix i_sem/journal inversion in mds_client_add (2333)
5900 - fix truncate/getattr lock cycle deadlock (2334)
5901 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5902 - fix two rare exit paths which could leak an l_lock() ref (2321)
5903 - fencepost error in MDS/OST orphan recovery (2226)
5904 - make log record alignment 8 bytes (1988)
5905 - lstripe now fails when requested offset > ost_count (2237)
5906 - ensure that all kernel series have a complete list.h (1607)
5907 - fix crashes in special-file operations (2316)
5908 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5910 - allow configurable automake binary, for testing new versions
5911 - small update to the lfs documentation
5913 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5915 * fix negative export reference count in fsfilt_sync (2312)
5917 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5918 * release candidate 0.9.1
5920 - orphans are moved into the PENDING directory for possible recovery
5921 - replayed opens now open by fid for orphan/rename safety (1042)
5922 - last close of an orphan inode generates a transno (683)
5923 - chdir() and mount() now pin the directory entry (1020)
5924 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5925 - discard very old requests without processing them (1502)
5926 - remove some common, well-understood CERRORs (1505)
5927 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5928 - clear "grant" flags in OST replies until OST grant code lands (1644)
5929 - fix read performance by not clobbering i_blksize on client (1598)
5930 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5931 - make LRU size a /proc tunable, clears locks when reduced (707)
5932 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5933 - server locks take references on exports now (1558)
5934 - build fixes for 2.4.20-rh trees (1663)
5935 - return an error from lov_create if all OSCs are inactive (1751)
5936 - fix import levels when a reconnect happens without a timeout (1597)
5937 - exit early from mds_open if we get a lookup error (1749)
5938 - partial page read at EOF wouldn't wait for disk before sending (1642)
5939 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5940 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5941 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5942 - write/truncate lock inversion (1639)
5943 - Don't auto-load obdclass, portals modules during cleanup (1495)
5944 - fix timestamps from jumping to "now" (1763)
5945 - extra journal assertions (1648)
5946 - add an extra multiunlink test (1771)
5947 - fix read_record/write_record API (1776)
5948 - fix leak of offset_extent, possible incorrect i_size later (1772)
5949 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5950 - Debugging for the kqswnal_get_idle_tx problems (1820)
5951 - Allow recovery to be attempted multiple times (1536)
5952 - Write out MDS last_rcvd file after it is first created (1600)
5953 - Fix tx_descriptor leak in failed transmit situations (1827)
5954 - ext3 journaling fixes for assertion failure after IO error (1871)
5955 - class_export_put() on freed export after completion AST error (1896)
5956 - Fix revalidate looping in VFS (1322)
5957 - Don't access a freed export during MDS_REINT timeout (1521)
5958 - Add open-unlink recovery support on the MDS (1673,1764)
5959 - Return an error if no MDS data was read from last_rcvd (1946)
5960 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5961 - Don't setup a disk device that doesn't match exported UUID (317)
5962 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5963 - avoid committing NULL handle in force close
5964 - local.sh is now a one-stripe LOV configuration
5965 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5966 - fix ext3 htree duplicate directory entry corruption (1516)
5967 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5968 - update to new LOV EA format (2097)
5969 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5971 2003-06-15 Phil Schwan <phil@clusterfs.com>
5974 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5975 - per-import/export recovery handling (958, 931, 959)
5976 - multiple last-rcvd slots, for serving multiple FSes (949)
5977 - connections are again shared between multiple imp/exports (963, 964)
5978 - "umount -f" would hang if any requests needed to be sent (393, 978)
5979 - avoid pinning large req buffer by copying for queued messages (989)
5980 - add "uuid" to "lctl device" command to help upcalls (991)
5981 - "open" RPCs with transnos would confuse recovery counters (1037)
5982 - do proper endian conversion of all wire messages (288, 340, 891)
5983 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5984 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5985 - fix ptlrpc_connection leak in target_handle_connect (1174)
5986 - fix import refcounting bug in OST and MDS cleanup (1134)
5987 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5988 - fix very unlikely obd_types race condition (501)
5989 - remove osc_open hack for echo_client (1187)
5990 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5991 - a failure in read_inode2 leads to deadlock (1139)
5992 - cancel ack-locks as soon as transaction is committed (1072)
5993 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5994 - make sure to commitrw after any preprw to avoid deadlock (1162)
5995 - failing to execute a file in a lustre FS would lock inode (1203)
5996 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5997 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5998 - obd_uuid2tgt would walk off the end of the list (1255)
5999 - on IA64 the osc would give portals incorrect bulk size (1258)
6000 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6001 - fix lock inversion caused by new llite matching code (1282)
6002 - limit the number of dirty pages on a client to 10MB (1286)
6003 - timed out locks were not being corrected cancelled (1289)
6004 - fix O_DIRECT above 4GB on IA-32 (1292)
6005 * major user-visible changes
6006 - fail out/fail over policy now controlled by the upcall (993)
6008 - add OBD_PING to check server availability and failure (954)
6009 - lustre messages are now sent in sending host order (288, 340, 891)
6010 - add eadatalen to MDS getattr reply (340)
6011 - OST read replies may contain second buffer, with per-page status (593)
6013 2003-03-11 Phil Schwan <phil@clusterfs.com>
6016 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6017 - failed open()s could cause deadlock; fixed (867, 869)
6018 - stop cancelling OST locks when files are closed (481)
6019 - overlapping XID spaces caused network corruption (851, 853)
6020 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6021 - setattr_raw added, to do single-RPC, server-side setattrs
6022 - lmc/lconf syntax change for OST UUIDs
6023 - fix crashy race condition between ptlrpc_free_req and osc_close
6024 - don't use request in mdc_enqueue if we hit a timeout (889)
6025 - don't set the inode i_size for regular files from the MDS (896)
6026 - handle out of order completion AST (842)
6027 - don't LBUG if a lock request times out after receiving AST (913)
6028 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6029 - if a bad lock AST arrives, send an error instead of dropping entirely
6030 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6031 - fix leak in bulk IO when only partially completed (899, 900, 926)
6032 - fix O_DIRECT for ia64 (55)
6033 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6034 - C-z after timeout could hang a process forever; fixed (977)
6036 - client-side I/O cache (678, 924, 929, 941, 970)
6038 - READPAGE and SETATTRs which don't take server-side locks get
6041 2003-02-11 Phil Schwan <phil@clusterfs.com>
6044 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6045 - Fix fsfilt_extN_readpage() to read a full page of directory
6046 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6047 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6048 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6049 - Most intent operations were removed, in favour of a new RPC mode
6050 that does a single RPC to the server and bypasses most of the VFS
6051 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6052 - Aggressively cancel local locks on DLM servers
6053 - mds_reint_unlink sends EA to the client if it's the last nlink.
6054 client uses that EA to unlink OST objects.
6055 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6056 - recursive symlinks were fixed (439)
6057 - fixed NULL deref in DEBUG_REQ
6058 - filter_update_lastobjid no longer calls sync, which annoyed extN
6059 - fixed multi-client small-writes to a single file problem (445)
6060 - fixed mtime updates during file writes (607)
6061 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6062 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6063 - fixed Linux OST size reporting problem (444, 656)
6064 - OST now updates object mtime with writes or setattr (607, 619)
6065 - client verifies file size before zeroing page past EOF (445)
6066 - OST now writes last allocated objid to disk with allocation (108)
6067 - LOV on echo now works (409)
6069 - mds_reint_unlink sends a new buffer, with the EA included. this
6070 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6071 set if a regular file was being unlinked, and it was the last link
6072 - use PtlGet from the target for bulk writes (315)
6073 - OST now updates object mtime with writes or setattr (607, 619)
6074 - LDLM now has a grant-time callback to revalidate locked items, if
6076 - Many MDS operations were reorganized to combat race conditions
6078 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6079 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6081 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6084 - Fully reactivate OST imports after reconnection (512, others)
6085 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6086 - More graceful error handling for truncating on dead OST (515)
6087 - Don't error out unless we're actually accessing dead stripes (474)
6088 - Fix garbage sizes when stripes are missing (410)
6089 - LRU counters were broken, causing constant lock purge (433, 432)
6090 - garbage on read from stripes with failed OSTs (441)
6091 - mark OSCs as active before reconnecting during recovery (438)
6092 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6093 - lfind did not preserve OST order in output (443)
6094 - symlinks cause hung clients, incorrect data (439)
6095 - stop dereferencing request after dropping refcount (457)
6096 - don't LASSERT(spin_is_locked) on non-SMP (455)
6097 - fixes for many rename() bugs
6098 - fstat didn't correctly synchronize attributes (399)
6099 - server must handle lock cancellation during blocking AST prep (487)
6100 - bulk descriptors were free()d too soon (511)
6101 - fix paths in lconf, which would load incorrect modules (451, 507)
6102 - fix confusing lconf 'host not found' error message (386)
6103 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6104 - fix race condition in mdc_blocking_ast() for inode access (526)
6105 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6106 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6107 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6108 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6109 - fix lstripe in several ways (kernel side) (527)
6110 - fix request leak in ldlm_cli_enqueue (262)
6111 - incorrect OSC was marked inactive after OST failure
6112 - call mds_fs_cleanup before unmounting filesystem (524)
6113 - fix races between taking ns_lock and ldlm_lock_change_resource
6114 - fix races updating LOV export open file list
6115 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6116 - fix recovery NULL deref in ldlm_cli_cancel_unused
6117 - fix some DLM races by using new hash table for lock handles (419)
6118 - permit the client to specify desired inodes, at replay
6119 - duplicate requests when we queue them for replay reintegration
6120 - fix last_rcvd offset calculation
6121 - sync after each recovered transaction, so we always make progress
6122 - never, not always, ERESTART requests without transnos
6123 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6125 - skip replay if the MDS says that the client is already connected
6126 - don't check for a recovery-enabled export to match lctl's UUID
6127 - don't INC_USE_COUNT for phantom exports
6128 - don't crash when cleaning up phantom exports (567)
6129 - don't double-finish or set replay data for errored mdc_open requests
6130 - abort requests when they time out, so we don't get old replies
6131 - send/receive replies for AST messages again
6132 - if the client says that it doesn't have the lock, cancel it on the
6134 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6135 mark it as destroyed, it will all work out when decref is called
6136 - fix module use counts (22, 581)
6138 - ASTs now expect a reply (server cancels lock on error reply)
6140 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6143 - fix many simultaneous client startup (392)
6144 - fix dentry->d_it clobbering
6145 - credentials weren't being shipped for readdir/getattr operations
6146 - remove invalid assertions triggered during some concurrent MD
6148 - proper Lustre versions added (336, 389)
6149 - fix memory leak for create error case (398)
6150 - fix LOV locking bug that would get cli/srv out of sync
6151 - fix echo client over LOV (409)
6152 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6153 - fix double-O_EXCL intent crash (424)
6154 - avoid sending multiple lock CANCELs (352)
6156 - MDS can do multi-client recovery (modulo bugs in new code)
6158 - many updates, edits, cleanups
6160 2002-11-18 Phil Schwan <phil@clusterfs.com>
6163 - fix null d_it dereference (346)
6164 - fix full OST/dbench hang (333)
6165 - fix permission problem with file removal (286)
6166 - fix removal of OSCs from LOV when they fail
6167 - fix NULL deref during bulk timeout (214)
6168 - fix problems related to multiple filesystems on one MDS (241)
6169 - fixed serious subtle metadata locking bugs
6170 - free locks on clients when inodes are removed due to memory
6172 - fix inode pointer in lock data (285)
6173 - partial support for multiple MDS on a single host (241)
6174 - data locks weren't cancelled at clear_inode time (290, 311)
6175 - intent locks could lead to unbounded lock growth (205)
6176 - added a maximum lock count, an LRU list, and a flusher
6177 - fix multiple rename (365)
6178 - properly abstracted the echo client
6179 - OSC locked 1 byte too many; fixed
6180 - rewrote brw callback code:
6181 - fixed recovery bugs related to LOVs (306)
6182 - fixed too-many-pages-in-one-write crash (191)
6183 - fixed (again) crash in sync_io_timeout (214)
6184 - probably fixed callback-related race (385)
6186 - Add capability to MDS protocol
6187 - LDLM cancellations and callbacks on different portals
6189 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6192 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6193 - defer open object destruction to close time (601981, 138)
6194 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6195 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6196 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6197 - don't call obd_setattr in ll_file_release for destroyed objects
6198 * protocol change to lustre_msg: move |version| and add |flags|
6199 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6200 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6201 * added replay of create, unlink, link and rename operations during
6202 MDS failover; recovery should be much more robust now
6203 * remove failed OSCs from LOVs (only lov_create uses this so far)
6204 * the lustre-HOWTO was brought (more) up to date (582544)
6206 2002-10-23 Phil Schwan <phil@clusterfs.com>
6209 - in-use dentries weren't being reused properly (617851)
6210 - prevent multiple LDLM setup (599178)
6211 - fix LOV size calculations for truncate (617853)
6212 - fix client handling of MDS intent errors (POSIX)
6213 - fix permission bug in lovstripe.c test (624321)
6214 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6215 - truncate past end of file could corrupt data
6216 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6217 - a race in recovery could return ETIMEDOUT to apps (623947)
6218 - building outside the source directory was fixed
6219 * the lustre-HOWTO was brought (more) up to date (582544)
6220 * major progress was made on recovery functionality
6222 2002-10-10 Phil Schwan <phil@clusterfs.com>
6225 - recovery deadlock fix
6226 - rm -rf causes LBUG fix (617817)
6227 - file open by multiple tasks fix (618962)
6228 - directory permissions bugs (602707 and 620007)
6229 - journal_stop fixed with locking (611313)
6230 - O_APPEND failures resolved (618273, perhaps 614459)
6231 - lconf PATH fix (619770)
6232 - IA64 build fix (621450)
6233 - RPC buffer sizes scale with amount of memory
6235 2002-10-01 Phil Schwan <phil@clusterfs.com>
6238 - locks would be cancelled without throwing away data pages,
6239 resulting in inconsistent data (605627)
6240 - inode attributes were not always being refreshed (605627, 612449)
6241 - lconf now continues to cleanup after lctl reports an error
6242 - MDS now enforces user permissions (602707)
6243 - lprocfs cleanup fixed, but not yet enabled (614157)
6244 - fixed infinite server hang, should a client not respond to an AST
6245 - avoid going into recovery if user calls readlink() with a buffer
6246 that's too small (613941)
6247 - AST RPCs no longer require replies (614867) -- this may be changed
6248 - don't crash server if client sends an IOV that's too big (611336)
6249 - fixed lock conversion deadlock (611892)
6250 - fixed the following of symlinks (614622)
6251 * recovery: the server can remove locks from a client that dies, other
6252 clients can make progress
6253 * more extN patch fixes
6254 * compile-time configurable ptlrpc buffer allocations
6256 - collaborative read cache document
6257 - Lustre Lite Performance CDR document-in-progress
6259 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6262 - fix typo in patch-2.4.18
6264 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6267 - clear ptlrpc request each time in handle_incoming_request()
6268 - unlink of files now destroys the object on the OST
6270 2002-09-19 Peter Braam <braam@clusterfs.com>
6272 * add hard link support
6273 * change obdfile creation method
6274 * kernel patch changed
6276 2002-09-19 Peter Braam <braam@clusterfs.com>
6279 - stack overflow bug in extN fixed
6281 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6283 * documentation updates
6284 - add man pages for config tools
6285 - update tests/README to describe testing with new config tools
6286 - finish metadata API descriptions
6287 * bug fixes and cleanups
6288 - statfs workaround for 16TB limit
6289 - LOV stripe allocation improved, can stripe on subset of OSTs
6290 - LOV file size/IO offset was wrong for files > 4GB in size
6291 - object EA data was being dropped, caused files to be unreadable
6292 - memory overflow with non-LOV OST caused memory corruption
6293 - fixed regression tests to work with new config tools, obdfilter
6294 - fixed bug when directory size became larger than 1 block
6295 - fixed bug (for single client case) when PWD was deleted
6296 - invalidate local directory pages when doing intent-based ops
6297 - avoid LDLM oops when lock callback contained bad data
6299 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6301 * documentation updates
6302 * bug fixes and cleanups
6303 - configuration tools
6306 - 64-bit compile warnings
6307 - 64-bit internal statfs data
6309 * test_brw on persistent OST devices
6311 * lprocfs (disabled)
6313 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6315 * documentation updates
6316 * bug fixes and cleanups
6317 * configuration tools
6319 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6322 * many small fixes to 0_5_4
6323 * io/network handling
6324 * thinkos in MDS operations
6326 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6329 * crucial basic fixes to 0.5.3
6330 * IOR, Iozone work over Elan
6333 2002-08-07 Phil Schwan <phil@clusterfs.com>
6334 * version 0_5_3, our first alpha
6335 * we use the new Portals iovs
6336 * documentation updates
6337 * bug fixes and cleanups
6338 * small changes in the DLM wire protocol
6340 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6341 * version 0_5_1 with some initial stability,
6342 * locking on MD and file I/O.
6343 * documentation updates
6344 * several bug fixes since 0.5.0
6345 * small changes in wire protocol
6347 2002-07-18 Phil Schwan <phil@clusterfs.com>
6349 * delivered as Lustre Light Alpha
6350 * fixed a crash after handling invalid MDS requests
6351 * fixed directory pages for architectures with non-4k pages sizes
6353 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6354 * release version v0_4_4
6355 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6357 * Fixes a number of interruption problems with OST operations.
6358 * Update documentation for portals header changes
6359 * Move all wire protocol structs/defines to lustre_idl.h
6360 * Fixes symlink length bug.
6361 * Add tcpdump to repository.
6363 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6364 * release version v0_4_3
6365 * Fixes statfs for inodes on extN.
6366 * Fixes bug in runtests which would delete /etc/hosts.
6367 * Use 64-bit object IDs wherever possible (not into VFS though)
6368 Remove ost_get_info, which is unused by lustre, and out of date.
6370 2002-07-03 Peter Braam <braam@clusterfs.com>
6371 * release version v0_4_2 Fixes a lookup error (type not passed)
6372 * move forward to head of Portals
6373 * move forward to latest Lustre kernel
6375 2002-06-25 Peter Braam <braam@clusterfs.com>
6376 * release version v0_4_1. Hopefully stable on single node use.